The Inference Report

Multi-shot video generation extends single-shot generation to coherent visual narratives, yet maintaining consistent characters, objects, and locations across shots remains a challenge over long sequences. Existing evaluations typically use independently generated prompt sets with limited entity coverage and simple consistency metrics, making standardized comparison difficult. We introduce EntityBench, a benchmark of 140 episodes (2,491 shots) derived from real narrative media, with explicit per-shot entity schedules tracking characters, objects, and locations simultaneously across easy / medium / hard tiers of up to 50 shots, 13 cross-shot characters, 8 cross-shot locations, 22 cross-shot objects, and recurrence gaps spanning up to 48 shots. It is paired with a three-pillar evaluation suite that disentangles intra-shot quality, prompt-following alignment, and cross-shot consistency, with a fidelity gate that admits only accurate entity appearances into cross-shot scoring. As a baseline, we propose EntityMem, a memory-augmented generation system that stores verified per-entity visual references in a persistent memory bank before generation begins. Experiments show that cross-shot entity consistency degrades sharply with recurrence distance in existing methods, and that explicit per-entity memory yields the highest character fidelity (Cohen's d = +2.33) and presence among methods evaluated. Code and data are available at https://github.com/Catherine-R-He/EntityBench/.

Visual reasoning, often interleaved with intermediate visual states, has emerged as a promising direction in the field. A straightforward approach is to directly generate images via unified models during reasoning, but this is computationally expensive and architecturally non-trivial. Recent alternatives include agentic reasoning through code or tool calls, and latent reasoning with learnable hidden embeddings. However, agentic methods incur context-switching latency from external execution, while latent methods lack task generalization and are difficult to train with autoregressive parallelization. To combine their strengths while mitigating their limitations, we propose ATLAS, a framework in which a single discrete 'word', termed as a functional token, serves both as an agentic operation and a latent visual reasoning unit. Each functional token is associated with an internalized visual operation, yet requires no visual supervision and remains a standard token in the tokenizer vocabulary, which can be generated via next-token prediction. This design avoids verbose intermediate visual content generation, while preserving compatibility with the vanilla scalable SFT and RL training, without architectural or methodological modifications. To further address the sparsity of functional tokens during RL, we introduce Latent-Anchored GRPO (LA-GRPO), which stabilizes the training by anchoring functional tokens with a statically weighted auxiliary objective, providing stronger gradient updates. Extensive experiments and analyses demonstrate that ATLAS achieves superior performance on challenging benchmarks while maintaining clear interpretability. We hope ATLAS offers a new paradigm inspiring future visual reasoning research.

Video generation powers a vast array of downstream applications. However, while the de facto standard, i.e., latent diffusion models, typically employ heavily conditioned denoising networks, their decoders often remain unconditional. We observe that this architectural asymmetry leads to significant loss of detail and inconsistency relative to the input image. To address this, we argue that the decoder requires equal conditioning to preserve structural integrity. We introduce RefDecoder, a reference-conditioned video VAE decoder by injecting high-fidelity reference image signal directly into the decoding process via reference attention. Specifically, a lightweight image encoder maps the reference frame into the detail-rich high-dimensional tokens, which are co-processed with the denoised video latent tokens at each decoder up-sampling stage. We demonstrate consistent improvements across several distinct decoder backbones (e.g., Wan 2.1 and VideoVAE+), achieving up to +2.1dB PSNR over the unconditional baselines on the Inter4K, WebVid, and Large Motion reconstruction benchmarks. Notably, RefDecoder can be directly swapped into existing video generation systems without additional fine-tuning, and we report across-the-board improvements in subject consistency, background consistency, and overall quality scores on the VBench I2V benchmark. Beyond I2V, RefDecoder generalizes well to a wide range of visual generation tasks such as style transfer and video editing refinement.

AI agents are being increasingly deployed in dynamic, open-ended environments that require adapting to new information as it arrives. To efficiently measure this capability for realistic use-cases, we propose building grounded simulations that replay real-world events in the order they occurred. We build FutureSim, where agents forecast world events beyond their knowledge cutoff while interacting with a chronological replay of the world: real news articles arriving and questions resolving over the simulated period. We evaluate frontier agents in their native harness, testing their ability to predict world events over a three-month period from January to March 2026. FutureSim reveals a clear separation in their capabilities, with the best agent's accuracy being 25%, and many having worse Brier skill score than making no prediction at all. Through careful ablations, we show how FutureSim offers a realistic setting to study emerging research directions like long-horizon test-time adaptation, search, memory, and reasoning about uncertainty. Overall, we hope our benchmark design paves the way to measure AI progress on open-ended adaptation spanning long time-horizons in the real world.

High-quality 3D scene reconstruction has recently advanced toward generalizable feed-forward architectures, enabling the generation of complex environments in a single forward pass. However, despite their strong performance in static scene perception, these models remain limited in responding to dynamic human instructions, which restricts their use in interactive applications. Existing editing methods typically rely on a 2D-lifting strategy, where individual views are edited independently and then lifted back into 3D space. This indirect pipeline often leads to blurry textures and inconsistent geometry, as 2D editors lack the spatial awareness required to preserve structure across viewpoints. To address these limitations, we propose VGGT-Edit, a feed-forward framework for text-conditioned native 3D scene editing. VGGT-Edit introduces depth-synchronized text injection to align semantic guidance with the backbone's spatial poses, ensuring stable instruction grounding. This semantic signal is then processed by a residual transformation head, which directly predicts 3D geometric displacements to deform the scene while preserving background stability. To ensure high-fidelity results, we supervise the framework with a multi-term objective function that enforces geometric accuracy and cross-view consistency. We also construct the DeltaScene Dataset, a large-scale dataset generated through an automated pipeline with 3D agreement filtering to ensure ground-truth quality. Experiments show that VGGT-Edit substantially outperforms 2D-lifting baselines, producing sharper object details, stronger multi-view consistency, and near-instant inference speed.

Generative video models are increasingly studied as implicit world models, yet evaluating whether they produce physically plausible 3D structure and motion remains challenging. Most existing video evaluation pipelines rely heavily on human judgment or learned graders, which can be subjective and weakly diagnostic for geometric failures. We introduce PDI-Bench (Perspective Distortion Index), a quantitative framework for auditing geometric coherence in generated videos. Given a generated clip, we obtain object-centric observations via segmentation and point tracking (e.g., SAM 2, MegaSaM, and CoTracker3), lift them to 3D world-space coordinates via monocular reconstruction, and compute a set of projective-geometry residuals capturing three failure dimensions: scale-depth alignment, 3D motion consistency, and 3D structural rigidity. To support systematic evaluation, we build PDI-Dataset, covering diverse scenarios designed to stress these geometric constraints. Across state-of-the-art video generators, PDI reveals consistent geometry-specific failure modes that are not captured by common perceptual metrics, and provides a diagnostic signal for progress toward physically grounded video generation and physical world model. Our code and dataset can be found at https://pdi-bench.github.io/.

Recent advances in Large Language Model (LLM) agents have enabled complex agentic workflows where models autonomously retrieve information, call tools, and reason over large corpora to complete tasks on behalf of users. Despite the growing adoption of retrieval-augmented generation (RAG) in agentic search systems, existing literature lacks a systematic comparison of how retrieval strategy choice interacts with agent architecture and tool-calling paradigm. Important practical dimensions, including how tool outputs are presented to the model and how performance changes when searches must cope with more irrelevant surrounding text, remain under-explored in agent loops. This paper reports an empirical study organized into two experiments. Experiment 1 compares grep and vector retrieval on a 116-question sample from LongMemEval, using a custom agent harness (Chronos) and provider-native CLI harnesses (Claude Code, Codex, and Gemini CLI), for both inline tool results and file-based tool results that the model reads separately. Experiment 2 compares grep-only and vector-only retrieval while progressively mixing in additional unrelated conversation history, so that each query is embedded in more distracting material alongside the passages that matter. Across Chronos and the provider CLIs, grep generally yields higher accuracy than vector retrieval in our comparisons in experiment 1; at the same time, overall scores still depend strongly on which harness and tool-calling style is used, even when the underlying conversation data are the same.

Mechanistic interpretability aims to break models into meaningful parts; verifying that two such parts implement the same computation is a prerequisite. Existing similarity measures evaluate either empirical behaviour, leaving them blind to out-of-distribution mechanisms, or basis-dependent parameters, meaning they disregard weight-space symmetries. To address these issues for the class of tensor-based models, we introduce a weight-based metric, tensor similarity, that is invariant to such symmetries. This metric captures global functional equivalence and accounts for cross-layer mechanisms using an efficient recursive algorithm. Empirically, tensor similarity tracks functional training dynamics, such as grokking and backdoor insertion, with higher fidelity than existing metrics. This reduces measuring similarity and verifying faithfulness into a solved algebraic problem rather than one of empirical approximation.

Scaling Scientific Machine Learning (SciML) toward universal foundation models is bottlenecked by negative transfer: the simultaneous co-training of disparate partial differential equation (PDE) regimes can induce gradient conflict, unstable optimization, and plasticity loss in dense neural operators. In particular, broadband open-channel fluid dynamics and boundary-dominated porous media flows impose incompatible spectral and geometric demands on a single dense parameter path. We introduce Shodh-MoE, a sparse-activated latent transformer architecture for multi-physics transport. Shodh-MoE operates on compressed 16^3 physical latents produced by a physics-informed autoencoder with an intra-tokenizer Helmholtz-style velocity parameterization, restricting decoded states to divergence-free velocity manifolds. The model guarantees exact mass conservation, achieving a physically verifiable velocity divergence of ~2.8 x 10^-10 (evaluated post-hoc in FP64) on 128^3 grids. A Top-1 soft-semantic router dynamically assigns localized latent patches to expert subnetworks, enabling specialized parameter paths for distinct physical mechanisms while preserving shared experts for universal symmetries. In a 20,000-step distributed pretraining run over mixed three-dimensional physical tensors, routing telemetry shows autonomous domain bifurcation: held-out validation tokens from the open-channel domain route exclusively to Expert 0, while porous-media tokens route exclusively to Expert 1. The model converges simultaneously across both regimes, achieving latent validation MSEs of 2.46 x 10^-5 and 9.76 x 10^-6, and decoded physical MSEs of 2.48 x 10^-6 and 1.76 x 10^-6. These results support sparse expert routing as a practical architectural mechanism for mitigating multi-physics interference in universal neural operators.

Test-time compute scaling is a primary axis for improving LLM reasoning. Existing methods primarily scale depth by extending a single reasoning trace. Scaling breadth by sampling multiple candidates in parallel is straightforward, but introduces a selection bottleneck: choosing the best candidate without a ground-truth verifier, since pointwise LLM judging is noisy and biased. To address this, we introduce OpenDeepThink, a population-based test-time compute framework that selects via pairwise Bradley-Terry comparison. Each generation, the LLM judges random pairs of candidates and aggregates votes via Bradley-Terry into a global ranking; top-ranked candidates are preserved and the top three quarters are mutated using the natural-language critiques produced during comparison; the bottom quarter is discarded. OpenDeepThink raises Gemini 3.1 Pro's effective Codeforces Elo by +405 points in eight sequential LLM-call rounds (~27 minutes wall-clock). The pipeline transfers across weaker and stronger models without retuning, and on the multi-domain HLE benchmark, gains appear concentrated in objectively verifiable domains and reverse in subjective ones. We release CF-73, a curated set of 73 expert-rated Codeforces problems with International Grandmaster annotation and 99% local-evaluation agreement against the official verdict.

Backdoor attacks pose a serious security threat to large language models (LLMs), which are increasingly deployed as general-purpose assistants in safety- and privacy-critical applications. Existing LLM backdoors rely primarily on content-based triggers, requiring explicit modification of the input text. In this work, we show that this assumption is unnecessary and limiting. We introduce MetaBackdoor, a new class of backdoor attacks that exploits positional information as the trigger, without modifying textual content. Our key insight is that Transformer-based LLMs necessarily encode token positions to process ordered sequences. As a result, length-correlated positional structure is reflected in the model's internal computation and can be used as an effective non-content trigger signal. We demonstrate that even a simple length-based positional trigger is sufficient to activate stealthy backdoors. Unlike prior attacks, MetaBackdoor operates on visibly and semantically clean inputs and enables qualitatively new capabilities. We show that a backdoored LLM can be induced to disclose sensitive internal information, including proprietary system prompts, once a length condition is satisfied. We further demonstrate a self-activation scenario, where normal multi-turn interaction can move the conversation context into the trigger region and induce malicious tool-call behavior without attacker-supplied trigger text. In addition, MetaBackdoor is orthogonal to content-based backdoors and can be composed with them to create more precise and harder-to-detect activation conditions. Our results expand the threat model of LLM backdoors by revealing positional encoding as a previously overlooked attack surface. This challenges defenses that focus on detecting suspicious text and highlights the need for new defense strategies that explicitly account for positional triggers in modern LLM architectures.

Disease screening is critical for early detection and timely intervention in clinical practice. However, most current screening models for medical images suffer from limited interpretability and suboptimal performance. They often lack effective mechanisms to reference historical cases or provide transparent reasoning pathways. To address these challenges, we introduce EviScreen, an evidential reasoning framework for disease screening that leverages region-level evidence from historical cases. The proposed EviScreen offers retrospection interpretability through regional evidence retrieved from dual knowledge banks. Using this evidential mechanism, the subsequent evidence-aware reasoning module makes predictions using both the current case and evidence from historical cases, thereby enhancing disease screening performance. Furthermore, rather than relying on post-hoc saliency maps, EviScreen enhances localization interpretability by leveraging abnormality maps derived from contrastive retrieval. Our method achieves superior performance on our carefully established benchmarks for real-world disease screening, yielding notably higher specificity at clinical-level recall. Code is publicly available at https://github.com/DopamineLcy/EviScreen.

Reconstructing precise clinical timelines is essential for modeling patient trajectories and forecasting risk in complex, heterogeneous conditions like sepsis. While unstructured clinical narratives offer semantically rich and contextually complete descriptions of a patient's course, they often lack temporal precision and contain ambiguous event timing. Conversely, structured electronic health record (EHR) data provides precise temporal anchors but misses a substantial portion of clinically meaningful events. We introduce a retrieval-augmented multimodal alignment framework that bridges this gap to improve the temporal precision of absolute clinical timelines extracted from text. Our approach formulates timeline reconstruction as a graph-based multistep process: it first extracts central anchor events from narratives to build an initial temporal scaffold, places non-central events relative to this backbone, and then calibrates the timeline using retrieved structured EHR rows as external temporal evidence. Evaluated using instruction-tuned large language models on the i2m4 benchmark spanning MIMIC-III and MIMIC-IV, our multimodal pipeline consistently improves absolute timestamp accuracy (AULTC) and improves temporal concordance across nearly all evaluated models over unimodal text-only reconstruction, without compromising event match rates. Furthermore, our empirical gap analysis reveals that 34.8% of text-derived events are entirely absent from tabular records, demonstrating that aligning these modalities can produce a more temporally faithful and clinically informative reconstruction of patient trajectories than either source alone.

This position paper argues that behavioural assurance, even when carefully designed, is being asked to carry safety claims it cannot verify. AI governance frameworks enacted between 2019 and early 2026 require reviewable evidence of properties such as the absence of hidden objectives, resistance to loss-of-control precursors, and bounded catastrophic capability; current assurance methodologies (primarily behavioural evaluations and red-teaming) are epistemically limited to observable model outputs and cannot verify the latent representations or long-horizon agentic behaviours these frameworks presume to regulate. We formalize this structural mismatch as the audit gap, the divergence between required and achievable verification access, and introduce the concept of fragile assurance to describe cases where the evidential structure does not support the asserted safety claim. Through an analysis of a 21-instrument inventory, we identify an incentive gradient where geopolitical and industrial pressures systematically reward surface-level behavioral proxies over deep structural verification. Finally, we propose a technical pivot: bounding the weight of behavioral evidence in legal text and extending voluntary pre-deployment access with mechanistic-evidence classes, specifically linear probes, activation patching, and before/after-training comparisons.

Vision-Language-Action (VLA) models are prone to compounding errors in dexterous manipulation, where high-dimensional action spaces and contact-rich dynamics amplify small policy deviations over long horizons. While Interactive Imitation Learning (IIL) can refine policies through human takeover data, applying it to high-degree-of-freedom (DoF) robotic hands remains challenging due to a command mismatch between human teleoperation and policy execution at the takeover moment, which causes abrupt robot-hand configuration changes, or "gesture jumps". We present Hand-in-the-Loop (HandITL), a seamless human-in-the-loop intervention method that blends human corrective intent with autonomous policy execution to avoid gesture jumps during bimanual dexterous manipulation. Compared with direct teleoperation takeover, HandITL reduces takeover jitter by 99.8% and preserves robust post-takeover manipulation, reducing grasp failures by 87.5% and mean completion time by 19.1%. We validate HandITL on tasks requiring bimanual coordination, tool use, and fine-grained long-horizon manipulation. When used to collect intervention data for policy refinement, HandITL yields policies that outperform those trained with standard teleoperation data by 19% on average across three long-horizon dexterous tasks.

Large language models (LLMs) achieve strong performance across a wide range of tasks, but remain frozen after pretraining until subsequent updates. Many real-world applications require timely, domain-specific information, motivating the need for efficient mechanisms to incorporate new knowledge. In this paper, we introduce MeMo (Memory as a Model), a modular framework that encodes new knowledge into a dedicated memory model while keeping the LLM parameters unchanged. Compared to existing methods, MeMo offers several advantages: (a) it captures complex cross-document relationships, (b) it is robust to retrieval noise, (c) it avoids catastrophic forgetting in the LLM, (d) it does not require access to the LLM's weights or output logits, enabling plug-and-play integration with both open and proprietary closed-source LLMs, and (e) its retrieval cost is independent of corpus size at inference time. Our experimental results on three benchmarks, BrowseComp-Plus, NarrativeQA, and MuSiQue, show that MeMo achieves strong performance compared to existing methods across diverse settings.

Reinforcement learning (RL) has emerged as a central paradigm for post-training LLM agents, yet its trajectory-level reward signal provides only coarse supervision for long-horizon interaction. On-Policy Self-Distillation (OPSD) complements RL by introducing dense token-level guidance from a teacher branch augmented with privileged context. However, transferring OPSD to multi-turn agents proves problematic: compounding multi-turn instability destabilizes supervision, while skill-conditioned privileged guidance requires asymmetric treatment for negative teacher rejections may arise from imperfect skills retrieval or utilization. We introduce SDAR (Self-Distilled Agentic Reinforcement Learning), which treats OPSD as a gated auxiliary objective while keeping RL as the primary optimization backbone. SDAR maps detached token-level signals into a sigmoid gate, strengthening distillation on teacher-endorsed positive-gap tokens and softly attenuating negative teacher rejections. Across the Qwen2.5 and Qwen3 families on ALFWorld, WebShop, and Search-QA, SDAR substantially improves over GRPO (+9.4% on ALFWorld, +7.0% on Search-QA, +10.2% on WebShop-Acc), avoids the instability of naive GRPO+OPSD, and consistently outperforms hybrid RL--OPSD baselines across model scales.

Feature attribution analysis is critical for interpreting machine learning models and supporting reliable data-driven decisions. However, feature attribution measures often exhibit stochastic variation: different train--test splits, random seeds, or model-fitting procedures can produce substantially different attribution values and feature rankings. This paper proposes a framework for incorporating stochastic nature of feature attribution and a robust attribution metric, RoSHAP, for stable feature ranking based on the SHAP metric. The proposed framework models the distribution of feature attribution scores and estimates it through bootstrap resampling and kernel density estimation. We show that, under mild regularity conditions, the aggregated feature attribution score is asymptotically Gaussian, which greatly reduces the computational cost of distribution estimation. The RoSHAP summarizes the distribution of SHAP into a robust feature-ranking criterion that simultaneously rewards features that are active, strong, and stable. Through simulations and real-data experiments, the proposed framework and RoSHAP outperform standard single-run attribution measures in identifying signal features. In addition, models built using RoSHAP-selected features achieve predictive performance comparable to full-feature models while using substantially fewer predictors. The proposed RoSHAP approach improves the stability and interpretability of machine learning models, enabling reliable and consistent insights for analysis.

We present Pelican-Unified 1.0, the first embodied foundation model trained according to the principle of unification. Pelican-Unified 1.0 uses a single VLM as a unified understanding module, mapping scenes, instructions, visual contexts, and action histories into a shared semantic space. The same VLM also serves as a unified reasoning module, autoregressively producing task-, action-, and future-oriented chains of thought in a single forward pass and projecting the final hidden state into a dense latent variable. A Unified Future Generator (UFG) then conditions on this latent variable and jointly generates future videos and future actions through two modality-specific output heads within the same denoising process. The language, video, and action losses are all backpropagated into the shared representation, enabling the model to jointly optimize understanding, reasoning, imagination, and action during training, rather than training three isolated expert systems. Experiments demonstrate that unification does not imply compromise. With a single checkpoint, Pelican-Unified 1.0 achieves strong performance across all three capabilities: 64.7 on eight VLM benchmarks, the best among comparable-scale models; 66.03 on WorldArena, ranking first; and 93.5 on RoboTwin, the second-best average among compared action methods. These results show that the unified paradigm succeeds in preserving specialist strength while bringing understanding, reasoning, imagination, and action into one model.

LLM quantization has become essential for memory-efficient deployment. Recent work has shown that quantization schemes can pose critical security risks: an adversary may release a model that appears benign in full precision but exhibits malicious behavior once quantized by users. However, existing quantization-conditioned attacks have been limited to relatively simple quantization methods, where the attacker can estimate weight regions that remain invariant under the target quantization. Notably, prior attacks have consistently failed to compromise more popular and sophisticated schemes, limiting their practical impact. In this work, we introduce the first quantization-conditioned attack that consistently induces malicious behavior that can be triggered by a broad range of advanced quantization techniques, including AWQ, GPTQ, and GGUF I-quants. Our attack exploits a simple property shared by many modern quantization methods: large outliers can cause other weights to be rounded to zero. Consequently, by injecting outliers into specific weight blocks, an adversary can therefore induce a targeted, predictable weight collapse in the model. This effect can be used to craft seemingly benign full-precision models that exhibit a wide range of malicious behaviors after quantization. Through extensive evaluation across three attack scenarios and LLMs, we show that our attack achieves high success rates against a broad range of quantization methods on which prior attacks fail. Our results demonstrate, for the first time, that the security risks of quantization are not restricted to simpler schemes but are broadly relevant across complex, widely-used quantization methods.

Standard unlearning evaluations measure behavioral suppression in full precision, immediately after training, despite every deployed language model being quantized first. Recent work has shown that 4-bit post-training quantization can reverse machine unlearning; we show this is not a tuning artefact but a systematic dual failure: gradient-based methods that achieve meaningful forgetting lose it under compression, while methods that survive quantization barely change the model. Both failures trace to the same root cause: across all baselines, per-parameter updates lie 47-828x below the NF4 quantization bin width; updates diffused across billions of parameters cannot clear quantization bin boundaries, a consequence we formalize as a sparsity-permanence tradeoff. We present MANSU (Mechanistic-Aligned Null-Space Unlearning), which resolves both modes by combining causal circuit attribution to isolate the minimal forget-set subgraph, circuit-restricted null-space projection with a diagonal-Fisher retain bound, and a per-parameter magnitude floor guaranteeing quantization survival by construction. We additionally introduce Circuit Attribution Divergence (CAD), a mechanistic verification metric distinguishing structural erasure from behavioral suppression, a distinction existing metrics cannot make. Across multiple model families and hazard benchmarks, MANSU is the first method to jointly satisfy all four properties with margin on each (meaningful forgetting, retain preservation, non-positive PTQ gap, and structural erasure), while gradient-based baselines recover up to +0.05 accuracy under compression.

Estimating how often an ML model will fail at deployment scale is central to pre-deployment safety assessment, but a feasible evaluation set is rarely large enough to observe the failures that matter. Jones et al. (2025) address this by extrapolating from the largest k failure scores in an evaluation set to predict deployment-scale failure rates. We give a finite-k decomposition of this estimator's forecast error and show that it has a built-in bias toward over-prediction in the typical case, which is the safety-favorable direction. This bias is offset when the evaluation set misses a rare high-failure mode that the deployment set contains, leaving the forecast to under-predict at deployment scale. We propose a fine-tuning objective, the forecastability loss, that addresses this failure mode. In two proof-of-concept experiments, a language-model password game and an RL gridworld, fine-tuning substantially reduces held-out forecast error while preserving primary-task capability and achieving safety similar to that of supervised baselines.

Causal inference, estimating causal effects from observational data, is a fundamental tool in many disciplines. Of particular importance across a variety of domains is the continuous treatment setting, where the variable of intervention has a continuous range. This setting is far less explored and represents a substantial shift from the binary treatment setting, with models needing to represent effects across a continuum of treatment values. In this paper, we present the first causal foundation model for the continuous treatment setting. Our model meta-learns the ability to predict causal effects across a wide variety of unseen tasks without additional training or fine-tuning. First, we design a novel prior over data-generating processes with continuous treatment variables in order to generate a rich causal training corpus. We then train a transformer to reconstruct individual treatment-response curves given only observational data, leveraging in-context learning to amortize expensive Bayesian posterior inference. Our model achieves state-of-the-art performance on individual treatment-response curve reconstruction tasks compared to causal models which are trained specifically for those tasks.

Autonomous multi-agent systems based on large language models (LLMs) have demonstrated remarkable abilities in independently solving complex tasks in a wide breadth of application domains. However, these systems hit critical reasoning, coordination, and computational scaling bottlenecks as the size and complexity of their tasks grow. These limitations hinder multi-agent systems from achieving high-throughput processing for highly parallelizable tasks, despite the availability of parallel computing and reasoning primitives in the underlying LLMs. We introduce the Agent-Parallel Workload Architecture (APWA), a distributed multi-agent system architecture designed for the efficient processing of heavily parallelizable agentic workloads. APWA facilitates parallel execution by decomposing workflows into non-interfering subproblems that can be processed using independent resources without cross-communication. It supports heterogeneous data and parallel processing patterns, and it accommodates tasks from a wide breadth of domains. In our evaluation, we demonstrate that APWA can dynamically decompose complex queries into parallelizable workflows and scales on larger tasks in settings where prior systems fail completely.

Reactive synthesis, the problem of automatically constructing a hardware circuit from a logical specification, is a long-standing challenge in formal verification. It is elusive for two reasons: It is algorithmically hard, and writing formal specifications by hand is notoriously difficult. In this paper, we tackle both sides of the problem. For the algorithmic side, we present a neuro-symbolic approach to reactive synthesis that couples large reasoning models with model checkers to iteratively repair a synthesized Verilog implementation via sound symbolic feedback. Our approach solves more benchmarks than the best dedicated tools in the annual synthesis competition and extends to constructing parameterized systems, a problem known to be undecidable. On the specification side, we introduce an autoformalization step that shifts the specification task from temporal logic to natural language by introducing a hand-authored dataset of natural-language specifications for evaluation. We demonstrate performance comparable to that of starting from formal specifications, establishing natural synthesis as a viable end-to-end workflow.

Long-term agent memory is increasingly multimodal, yet existing evaluations rarely test whether agents preserve the visual evidence needed for later reasoning. In prior work, many visually grounded questions can be answered using only captions or textual traces, allowing answers to be inferred without preserving the fine-grained visual evidence. Meanwhile, harder cases that require reasoning over changing visual states are largely absent. Therefore, we introduce MemEye, a framework that evaluates memory capabilities from two dimensions: one measures the granularity of decisive visual evidence (from scene-level to pixel-level evidence), and the other measures how retrieved evidence must be used (from single evidence to evolutionary synthesis). Under this framework, we construct a new benchmark across 8 life-scenario tasks, with ablation-driven validation gates for assessing answerability, shortcut resistance, visual necessity, and reasoning structure. By evaluating 13 memory methods across 4 VLM backbones, we show that current architectures still struggle to preserve fine-grained visual details and reason about state changes over time. Our findings show that long-term multimodal memory depends on evidence routing, temporal tracking, and detail extraction.

Moving to a new culture and adapting to a new life, as an international student, can be a stressful experience. In the US, international students face unique overlapping challenges, yet the current support ecosystem, including university support systems and informal social networks, remains largely fragmented. While conversational AI has emerged as a tool used by many (e.g., generative AI chatbots like ChatGPT and Google Gemini), we do not have a clear understanding of how international students adopt and perceive these technologies as support tools. We conducted a survey study (n=60) to map the relationship between international students' challenges and AI adoption patterns, followed by an interview study with 14 participants to identify the underlying motivations and boundaries of use. Our findings show that AI is perceived as a first-aid tool for immediate challenges, however, there is an interest in transforming AI from a tool for short-term help into a long-term support companion. By identifying where and how AI can provide long-term support, and where it is insufficient, we contribute recommendations for creating AI-powered support tailored to the unique needs of international students.

Robust state estimation for highly dynamic motion of legged robots remains challenging, especially in dynamic, contact-rich scenarios. Traditional approaches often rely on binary contact states that fail to capture the nuances of partial contact or directional slippage. This paper presents CoCo-InEKF, a differentiable invariant extended Kalman filter that utilizes continuous contact velocity covariances instead of binary contact states. These learned covariances allow the method to dynamically modulate contact confidence, accounting for more nuanced conditions ranging from firm contact to directional slippage or no contact. To predict these covariances for a set of predefined contact candidate points, we employ a lightweight neural network trained end-to-end using a state-error loss. This approach eliminates the need for heuristic ground-truth contact labels. In addition, we propose an automated contact candidate selection procedure and demonstrate that our method is insensitive to their exact placement. Experiments on a bipedal robot demonstrate a superior accuracy-efficiency tradeoff for linear velocity estimation, as well as improved filter consistency compared to baseline methods. This enables the robust execution of challenging motions, including dancing and complex ground interactions -- both in simulation and in the real world.

End-to-end autonomous driving planners are commonly trained by imitating a single logged trajectory, yet evaluated by rule-based planning metrics that measure safety, feasibility, progress, and comfort. This creates a training--evaluation mismatch: trajectories close to the logged path may violate planning rules, while alternatives farther from the demonstration can remain valid and high-scoring. The mismatch is especially limiting for proposal-selection planners, whose performance depends on candidate-set coverage and scorer ranking quality. We propose CLOVER, a Closed-LOop Value Estimation and Ranking framework for end-to-end autonomous driving planning. CLOVER follows a lightweight generator--scorer formulation: a generator produces diverse candidate trajectories, and a scorer predicts planning-metric sub-scores to rank them at inference time. To expand proposal support beyond single-trajectory imitation, CLOVER constructs evaluator-filtered pseudo-expert trajectories and trains the generator with set-level coverage supervision. It then performs conservative closed-loop self-distillation: the scorer is fitted to true evaluator sub-scores on generated proposals, while the generator is refined toward teacher-selected top-$k$ and vector-Pareto targets with stability regularization. We analyze when an imperfect scorer can improve the generator, showing that scorer-mediated refinement is reliable when scorer-selected targets are enriched under the true evaluator and updates remain conservative. On NAVSIM, CLOVER achieves 94.5 PDMS and 90.4 EPDMS, establishing a new state of the art. On the more challenging NavHard split, it obtains 48.3 EPDMS, matching the strongest reported result. On supplementary nuScenes open-loop evaluation, CLOVER achieves the lowest L2 error and collision rate among compared methods. Code data will be released at https://github.com/WilliamXuanYu/CLOVER.

We introduce a reusable framework for auditing whether LLM attack benchmarks collectively cover the threat surface: a 4$\times$6 Target $\times$ Technique matrix grounded in STRIDE, constructed from a 507-leaf taxonomy -- 401 data-populated and 106 threat-model-derived leaves -- of inference-time attacks extracted from 932 arXiv security studies (2023--2026). The matrix enables benchmark-external validation -- auditing collective coverage rather than individual benchmark consistency. Applying it to six public benchmarks reveals that the three primary frameworks (HarmBench, InjecAgent, AgentDojo) occupy non-overlapping cells covering at most 25\% of the matrix, while entire STRIDE threat categories (Service Disruption, Model Internals) lack any standardized evaluation, despite published attacks in these categories achieving 46$\times$ token amplification and 96\% attack success rates through mechanisms which no benchmark tests. The corpus of 2,521 unique attack groups further reveals pervasive naming fragmentation (up to 29 surface forms for a single attack) and heavy concentration in Safety \& Alignment Bypass, structural properties invisible at smaller scale. The taxonomy, attack records, and coverage mappings are released as extensible artifacts; as new benchmarks emerge, they can be mapped onto the same matrix, enabling the community to track whether evaluation gaps are closing.

Reinforcement learning from verifiable rewards (RLVR) suffers from sparse outcome signals, creating severe exploration bottlenecks on complex reasoning tasks. Recent on-policy self-distillation methods attempt to address this by utilizing language feedback to generate dense, token-level supervision. However, these approaches rely on a fixed, passive teacher to interpret the feedback. As the student policy improves, the teacher's zero-shot assessment capabilities plateau, ultimately halting further learning. To overcome this, we propose Variational Policy Distillation (VPD), a framework that formalizes learning from language feedback as a Variational Expectation-Maximization (EM) problem. VPD co-evolves both policies: in the E-step, the teacher is actively refined on trajectory outcomes via an adaptive trust-region update, translating textual feedback into a dynamically improved target token distribution. In the M-step, the student internalizes this dense distributional guidance on its own on-policy rollouts. By continuously improving the teacher's ability to extract actionable signals from textual critique, VPD overcomes the limitations of passive distillation. Evaluated across diverse sources of diagnostic feedback on scientific reasoning and code generation tasks, VPD consistently outperforms both standard RLVR and existing self-distillation baselines. Finally, by stress-testing our framework on rigid mathematical reasoning and cold-start regimes, we illuminate the fundamental bounds of feedback-driven self-distillation compared to pure environment-driven RL.

Adaptations of features commonly applied in the field of visual computing, co-occurrence matrix (COM) and run-length matrix (RLM), are proposed for the similarity computation of strings in general (words, phrases, codes and texts). The proposed features are not sensitive to language related information. These are purely statistical and can be used in any context with any language or grammatical structure. Other statistical measures that are commonly employed in the field such as longest common subsequence, maximal consecutive longest common subsequence, mutual information and edit distances are evaluated and compared. In the first synthetic set of experiments, the COM and RLM features outperform the remaining state-of-the-art statistical features. In 3 out of 4 cases, the RLM and COM features were statistically more significant than the second best group based on distances (P-value < 0.001). When it comes to a real text plagiarism dataset, the RLM features obtained the best results.

Retrieval-Augmented Generation can improve factuality by grounding answers in external evidence, but Agentic GraphRAG complicates what it means for citations to be faithful. In these systems, an agent explores a knowledge graph before producing an answer and a small set of citations. We frame citation faithfulness as a trajectory-level problem: final citations should not only support the answer, but also account for the graph traversal, structure, and visited-but-uncited entities that may influence it. Through controlled ablation experiments, we compare the effects of isolating, removing, and masking cited and uncited graph entities. Our results show that cited evidence is often necessary, as removing it substantially changes answers and reduces accuracy. However, citations are not sufficient, because accurate answers can also depend on uncited traversal context and surrounding graph structure. These findings suggest that citation evaluation in Agentic GraphRAG should move beyond source support toward provenance over the broader retrieval trajectory.

Off-policy evaluation (OPE) estimates the value of a target treatment policy (e.g., a recommender system) using data collected by a different logging policy. It enables high-stakes experimentation without live deployment, yet in practice accuracy depends heavily on the logging policy used to collect data for computing the estimate. We study how to design logging policies that minimize OPE error for given target policies. We characterize a fundamental reward-coverage tradeoff: concentrating probability mass on high-reward actions reduces variance but risks missing signal on actions the target policy may take. We propose a unifying framework for logging policy design and derive optimal policies in canonical informational regimes where the target policy and reward distribution are (i) known, (ii) unknown, and (iii) partially known through priors or noisy estimates at logging time. Our results provide actionable guidance for firms choosing among multiple candidate recommendation systems. We demonstrate the importance of treatment selection when gathering data for OPE, and describe theoretically optimal approaches when this is a firm's primary objective. We also distill practical design principles for selecting logging policies when operational constraints prevent implementing the theoretical optimum.

Voice agents increasingly require reliable tool use from speech, whereas prominent tool-calling benchmarks remain text-based. We study whether verified text benchmarks can be converted into controlled audio-based tool calling evaluations without re-annotating the tool schema and gold labels. Our dataset-agnostic framework uses text-to-speech, speaker variation, and environmental noise to create paired text-audio instances while preserving the original dataset annotations. Based on extensive evaluation of 7 omni-modal models on audio-converted versions of Confetti and When2Call, our framework demonstrates that the performance is strongly model- and task-dependent: Gemini-3.1-Flash-Live obtains the highest Confetti score (70.4), whereas GPT-Realtime-1.5 performs best on When2Call (71.9). On Confetti, the text-to-voice gap ranges from 1.8 points for Qwen3-Omni to 4.8 points for GPT-Realtime-1.5. A targeted analysis of failure cases demonstrates that degradations most often reflect misunderstandings of argument values in the speech. Considering real-world deployment scenarios, we further report text-only results, an ambiguity-based reformulation stress test, and a reference-free LLM-as-judge protocol validated against human preferences. Notably, we find that open-source Qwen3 judges with at least 8B parameters exceed 80% agreement with proprietary judges, supporting privacy-preserving evaluation. Overall, our framework provides a verifiable and reproducible first-stage diagnostic that complements purpose-built audio corpora.

Large language model (LLM) based multi-turn dialogue systems often struggle to track dependencies across non-adjacent turns, undermining both consistency and scalability. As conversations lengthen, essential information becomes sparse and is buried in irrelevant context, while processing the entire dialogue history incurs severe efficiency bottlenecks. Existing solutions either rely on high latency external memory or lose fine-grained details through iterative summarization. In this paper, we propose Self-Recall Thinking (SRT), a framework designed to address long-range contextual dependency and sparse informative signals in multi-turn dialogue. SRT identifies helpful historical turns and uses them to generate contextually appropriate responses, enabling the model to selectively recall and reason over context during inference. This process yields an endogenous reasoning process that integrates interpretable recall steps without external modules. SRT incorporates: (1) Dependency Construction: Generating and converting it into self-recall chains; (2)Capability Initialization: Training to enable reasoning chains with recall tokens capability; (3)Reasoning Improvement: Refining accuracy via verifiable rewards to optimize recall and reasoning for correct answers. Experiments on multiple datasets demonstrate that SRT improves F1 score by 4.7% and reduces end-to-end latency by 14.7% over prior methods, achieving a balance between reasoning latency and accuracy, and outperforming state-of-the-art baselines.

Large Language Models (LLMs) have demonstrated remarkable abilities in reasoning. However, maximizing their potential through inference-time scaling faces challenges in trade-off between sampling budget and reasoning quality. Current strategies remain inefficient as they typically treat sampling width and depth as orthogonal objectives, where width consensus methods risk reinforcing hallucinations, while depth pruning mechanisms prematurely truncate complex yet valid reasoning chains. Therefore, we propose Dual-Dimensional Consistency (DDC), a unified framework that bridges path quality with adaptive termination. By coupling Confidence-Weighted Bayesian protocol with a Trend-Aware Stratified Pruning, our method ensures that computational resources are concentrated on high quality reasoning paths, filtering hallucinations while accelerating consensus. Evaluations across five benchmarks demonstrate that this approach reduces token consumption by over 10 times while maintaining or exceeding the accuracy of strong baselines across various LLMs.

Detecting memory corruption vulnerabilities in stripped binaries requires recovering object semantics, interprocedural propagation, and feasible triggers from low-level, lossy representations. Recent LLM-based approaches improve code understanding, but reliable detection still requires grounding in memory-relevant semantics and runtime feasibility evidence. We present Veritas, a semantically grounded framework for binary memory corruption vulnerability detection. Veritas combines a static slicer over RetDec-lifted LLVM IR, a dual-view LLM detector that reasons step by step over grounded flows using decompiled C and selective LLVM IR, and a multi-agent validator that checks hypotheses against debugger-visible artifacts and runtime evidence. The slicer reconstructs value-flow relations from LLVM-IR facts, including def-use, calls, returns, globals, and pointer operations, and emits compact witness-backed flow objects. The detector uses these artifacts to reason about control flow, bounds, and object correspondence without rediscovering whole-binary propagation. The validator confirms or rejects candidates through guided debugging, breakpoint inspection, and memory-checking oracles. We implement Veritas as a modular pipeline and evaluate it on a curated benchmark of real-world binary vulnerability cases. Veritas achieves 90\% recall. For false-positive assessment, we exhaustively validate and manually verify 623 detector candidates and audit additional candidates from larger cases. The exhaustive subset produces no false positives, while the additional audit identifies two confirmed false positives. In a real-world application, Veritas discovered a previously unknown Apple vulnerability that was confirmed and assigned a CVE. These results support semantic grounding as an operational design principle for practical binary vulnerability detection.

Nowadays refinery optimization utilizes sheer amounts of data, which can be handled with modern Linear Programming (LP) software, but the interpreting and applying the results remains challenging. Large petrochemical companies use massive models, with hundreds of thousands of input matrix elements. The LP solution is mathematically correct, but simplifications are made in the model, and data supply errors may occur. Therefore, further insight is needed to trust the results. The LP solver does not have a memory, so additional understanding could be gained by analyzing historical data and comparing it to the current plan. As such, machine learning approaches were suggested to support decision making based on the LP solution. Among these, Anomaly Detection tools are proposed to be used in tandem with the LP output. A transformed version of the popular ECOD methodology is applied. New methods are proposed to handle high-dimensional data: choosing the most informative pairs. Then, this is used alongside two 2D Anomaly Detection algorithms, revealing several business opportunities and data supply errors in the MOL refinery scheduling and planning architecture.

The choice of optimiser is important in deep learning, as it strongly influences model efficiency and speed of convergence. However, many commonly used optimisers encounter difficulties when applied to imbalanced and sequential datasets, limiting their ability to capture patterns of minority classes. In this study, we propose Dynamic Batch-Sensitive Adam (DBS-Adam), an optimiser that dynamically scales the learning rate using a batch difficulty score derived from exponential moving averages of gradient norms and batch loss. DBS-Adam improves training stability and accelerates convergence by increasing updates for difficult batches and reducing them for easier ones. We evaluate DBS-Adam by integrating it with Bi-Directional LSTM networks for accident injury severity prediction, addressing class imbalance through SMOTE-ENN resampling and Focal Loss. Four experimental configurations compare baseline Bi-LSTM models and alternative architectures to assess optimiser impact. Rigorous comparison against state-of-the-art optimisers (AMSGrad, AdamW, AdaBound) across five random seeds demonstrated DBS-Adam's competitive performance with statistically significant precision improvements (p=0.020). Results indicate that DBS-Adam outperforms standard optimisation approaches, achieving 95.22% test accuracy, 96.11% precision, 95.28% recall, 95.39% F1-score, and a test loss of 0.0086. The proposed framework enables effective real-time accident severity classification for targeted emergency response and road safety interventions, demonstrating the value of DBS-Adam for learning from imbalanced sequential data.

We study a prototypical situation when a learned predictor can discover useful low-dimensional structure in data, while using fewer samples than are needed for accurate prediction. Specifically, we consider the problem of recovering a multi-index polynomial $f^*(x)=h(Ux)$, with $U\in\mathbb{R}^{r\times d}$ and $r\ll d$, from finitely many data/label pairs. Importantly, the target function depends on input $x$ only through the projection onto an unknown $r$-dimensional central subspace. The algorithm we analyze is appealingly simple: fit kernel ridge regression (KRR) to the data and compute the Average Gradient Outer Product (AGOP) from the fitted predictor. Our main results show that under reasonable assumptions the top $r$-dimensional eigenspace of AGOP provably recovers the central subspace, even in regimes when the prediction error remains large. Specifically, if the target function $f^*$ has degree $p^*$, it is known that $n\asymp d^{p^*}$ samples are necessary for KRR to achieve accurate prediction. In contrast, we show that if a low degree $p$ component of $f^*$ already carries all relevant directions for prediction, subspace recovery occurs in the much lower sample regime $n\asymp d^{p+δ}$ for any $δ\in(0,1)$. Our results thus demonstrate a separation between prediction and representation, and provide an explanation for why iterative kernel methods such as Recursive Feature Machines (RFM) can be sample-efficient in practice.

The development of high-quality text embeddings is increasingly drifting toward an exclusionary future, defined by three critical barriers: prohibitive computational costs, a narrow linguistic focus that neglects most of the world's languages, and a lack of transparency from closed-source or open-weight models that stifles research. To dismantle these barriers, we introduce ML-Embed, a suite of inclusive and efficient models built upon a new framework: 3-Dimensional Matryoshka Learning (3D-ML). Our framework addresses the computational challenge with comprehensive efficiency across the entire model lifecycle. Beyond the storage benefits of Matryoshka Representation Learning (MRL) and flexible inference-time depth provided by Matryoshka Layer Learning (MLL), we introduce Matryoshka Embedding Learning (MEL) for enhanced parameter efficiency. To address the linguistic challenge, we curate a massively multilingual dataset and train a suite of models ranging from 140M to 8B parameters. In a direct commitment to transparency, we release all models, data, and code. Extensive evaluation on 430 tasks demonstrates that our models set new records on 9 of 17 evaluated MTEB benchmarks, with particularly strong results in low-resource languages, providing a reproducible blueprint for building globally equitable and computationally efficient AI systems.

Croissant has emerged as the metadata standard for machine learning datasets, providing a structured, JSON-LD-based format that makes dataset discovery, automated ingestion, and reproducible analysis machine-checkable across ML platforms. Adoption has accelerated, and NeurIPS now requires Croissant metadata in every submission to its dataset tracks. Yet in practice Croissant generation usually starts with uploading data to a public platform, a path infeasible for governed and large local repositories that hold much of the high-value data ML increasingly relies on. We release Croissant Baker, a local-first, open-source command-line tool that generates validated Croissant metadata directly from a dataset directory through a modular handler registry. We evaluate Croissant Baker on over 140 datasets, scaling to MIMIC-IV at 886 million rows and 374 Parquet files. On held-out comparisons against producer-authored or standards-derived ground truth, Croissant Baker reaches 97-100% agreement across multiple domains.

Function calling, also known as tool use, is a core capability of modern LLM agents but is typically constrained by synchronous execution semantics. Under these semantics, LLM decoding is blocked until each function call completes, resulting in increasing end-to-end latency. In this work, we introduce AsyncFC, a pure execution-layer framework that decouples LLM decoding from function execution, enabling overlap between model decoding and function execution as well as inter-function parallelism when dependencies permit. AsyncFC layers over existing models and unmodified function implementations, requiring no fine-tuning or changes to the standard synchronous function-calling protocol. Across standard function-calling benchmarks and adapted software engineering benchmarks, AsyncFC significantly reduces end-to-end task completion time while preserving task accuracy. Furthermore, these results reveal that LLMs possess a native capability to reason over symbolic futures that represent unresolved execution results, enabling an asynchronous paradigm for model-tool interaction.

Vision-Language Models (VLMs) are increasingly applied to cultural heritage materials, from digital archives to educational platforms. This work identifies a fundamental issue in how these models interpret historical artifacts. We define this phenomenon as cultural anachronism, the tendency to misinterpret historical objects using temporally inappropriate concepts, materials, or cultural frameworks. To quantify this phenomenon, we introduce the Temporal Anachronism Benchmark for Vision-Language Models (TAB-VLM), a dataset of 600 questions across six categories, designed to evaluate temporal reasoning on 1,600 Indian cultural artifacts spanning prehistoric to modern periods. Systematic evaluations of ten state-of-the-art models reveal significant deficiencies on our benchmark, and even the best model (GPT-5.2) achieves only 58.7% overall accuracy. The performance gap persists across varying architectures and scales, suggesting that cultural anachronism represents a significant limitation in visual AI systems, regardless of model size. These findings highlight the disparity between current VLM capabilities and the requirements for accurately interpreting cultural heritage materials, particularly for non-Western visual cultures underrepresented in training data. Our benchmark provides a foundation for enhancing temporal cognition in multimodal AI systems that interact with historical artifacts. The dataset and code are available in our project page.

The rapid expansion of spiking neural networks (SNNs) has led to a proliferation of training algorithms that differ widely in biological inspiration, computational structure, and hardware suitability. Despite this progress, the field lacks a unified, fine-grained taxonomy that systematically organizes these approaches and clarifies their conceptual relationships. This survey provides a comprehensive taxonomy of SNN training algorithms, spanning surrogate-gradient backpropagation, local and three-factor learning rules, biologically inspired plasticity mechanisms, ANN-to-SNN conversion pipelines, and non-standard optimization strategies. We analyze each class in terms of its computational principles, learning signals, and locality properties. To support reproducible research, we release NeuroTrain, an open-source snnTorch-based framework that implements a representative set of these algorithms within a unified, modular, and extendable framework, enabling consistent benchmarking across datasets, architectures, and training regimes. By consolidating fragmented literature and providing a reusable benchmarking framework, this survey identifies common patterns, highlights open challenges, and outlines promising directions for future work on scalable, efficient SNN training.

Reinforcement learning has emerged as a powerful tool for improving diffusion-based text-to-image models, but existing methods are largely limited to single-task optimization. Extending RL to multiple tasks is challenging: joint optimization suffers from cross-task interference and imbalance, while cascade RL is cumbersome and prone to catastrophic forgetting. We propose DiffusionOPD, a new multi-task training paradigm for diffusion models based on Online Policy Distillation (OPD). DiffusionOPD first trains task-specific teachers independently, then distills their capabilities into a unified student along the student own rollout trajectories. This decouples single-task exploration from multi-task integration and avoids the optimization burden of solving all tasks jointly from scratch. Theoretically, we lift the OPD framework from discrete tokens to continuous-state Markov processes, deriving a closed-form per-step KL objective that unifies both stochastic SDE and deterministic ODE refinement via mean-matching. We formally and empirically demonstrate that this analytic gradient provides lower variance and better generality compared to conventional PPO-style policy gradients. Extensive experiments show that DiffusionOPD consistently surpasses both multi-reward RL and cascade RL baselines in training efficiency and final performance, while achieving state-of-the-art results on all evaluated benchmarks.

Continually pre-training a large language model on heterogeneous text domains, without replay or task labels, has remained an unsolved architectural problem at LLM scale. Existing methods rely on replay buffers, task identifiers, regularization penalties that scale poorly, or sentence-classification-scale evaluation. We introduce TFGN, an architectural overlay for transformer language models that produces input-conditioned, parameter-efficient updates while leaving the rest of the transformer unchanged. On six heterogeneous text domains (Prose, Python, Math, Biomedical, Chinese, JavaScript) at 1B tokens per phase across three model scales (~398M, ~739M, ~9B) and two regimes (From-Scratch and Retrofit), TFGN achieves backward transfer of -0.007 at LLaMA 3.1 8B Retrofit, HellaSwag retention 0.506/0.504/0.510, and >=99.59% L2-orthogonal gradient separation between domain pairs - with no replay, no task IDs, no Fisher penalty. The same matrices show positive cross-domain forward transfer: held-out JavaScript PPL drops 26.8% at LLaMA-8B Retrofit and 62.0% at GPT-2 Medium From-Scratch purely from Python training. Two extensions on the same substrate close further open problems. A closed-loop meta-control layer (Extension A) reduces forgetting by an additional 81% at ~398M, mapping onto the System A and System M roles of Dupoux et al. (arXiv:2603.15381). An operator-level plan vector (Extension B) reshapes forward-pass behavior at 99.96% cosine fidelity over 30 source->target pairs. The architectural insight is a Read/Write decomposition: the forward pass is fully dense, while cross-domain parameter updates are structured so prior-domain subspaces are not written to. To our knowledge, TFGN is the first architecture that simultaneously closes catastrophic forgetting at LLM scale, realizes a closed-loop autonomous-learning meta-controller, and carries an operator-level latent planner.

Speculative decoding (SD) accelerates large language model (LLM) inference by using a smaller draft model to propose multiple tokens that are verified by a larger target model in parallel. While prior work demonstrates substantial speedups in isolated or fixed-batch settings, the behavior of SD in production serving systems remains poorly understood: request load varies over time, and effective batch size emerges from the serving system rather than being directly controlled or observed. In this work, we develop a simple and interpretable latency model for SD in LLM serving. We infer effective batch size from request rate using Little's Law and decompose per-request demand into load-independent and load-dependent components for prefill, drafting, and verification. We validate our model using extensive measurements from vLLM across verifier and drafter model sizes, prefill and decode lengths, request rates, draft lengths, and acceptance probabilities. The model accurately describes observed latency, explains why speedups often diminish as server load increases, and characterizes how draft length, acceptance rate, and verifier-drafter size shape latency across serving conditions, with implications for configuring SD in deployed systems. We further show how the framework extends to mixture of experts models, where sparse expert activation changes the effective service costs across load regimes. Together, our results provide a structured framework for understanding SD in real LLM serving systems.

Recently, deep generative models have been used for posterior inference in inverse problems, including high-stakes applications in medical imaging and scientific discovery, where the uncertainty of a prediction can matter as much as the prediction itself. However, posterior uncertainty is difficult to interpret because it can mix ambiguity inherent to the forward operator with uncertainty propagated through inference. We introduce a structural decomposition of posterior uncertainty that isolates intrinsic ambiguity. A cascade formulation makes this ambiguity accessible for calibration analysis, enabling qualitative diagnostics and simulation-based calibration tests that reveal failure modes that remain hidden when models are selected by reconstruction quality alone. We first validate the approach on a Gaussian example with analytical posterior structure, then illustrate the decomposition on accelerated magnetic resonance imaging (MRI), and finally apply the calibration diagnostics to electroencephalography (EEG) source imaging.

As audio-first agents become increasingly common in physical AI, conversational robots, and screenless wearables, audio large language models (audio-LLMs) must integrate speaker-specific understanding to support user authorization, personalization, and context-aware interaction. This requires modeling who is speaking, how the voice sounds, and how recording conditions affect speaker cues. Conventional speaker verification systems provide strong scalar scores but little linguistic evidence, while current audio-LLMs and speaker-aware language models have limited ability to organize speaker information beyond binary labels or descriptive profiles. We present SpeakerLLM, a speaker-specialized audio-LLM framework that unifies single-utterance speaker profiling, recording-condition understanding, utterance-pair speaker comparison, and evidence-organized verification reasoning within a natural-language interface. We construct verification-reasoning targets and a decision-composition policy that separate profile-level evidence from the final same-or-different decision and organize recording condition, profile evidence, and the decision into a structured trace. At its core, SpeakerLLM uses a hierarchical speaker tokenizer designed to capture multiple granularities of speaker evidence. Utterance-level speaker embeddings summarize identity and profile-level cues, whereas frame-level speaker features preserve fine-grained acoustic descriptors. Experiments show that SpeakerLLM-Base improves speaker-profile and recording-condition understanding over general audio-LLMs, while SpeakerLLM-VR preserves strong generated-verdict accuracy and produces decision traces grounded in the supervised verification reasoning schema. We will release the metadata-enriched supervision dataset and target-construction code for reproducibility.

We propose EverAnimate, an efficient post-training method for long-horizon animated video generation that preserves visual quality and character identity. Long-form animation remains challenging because highly dynamic human motion must be synthesized against relatively static environments, making chunk-based generation prone to accumulated drift: (i) low-level quality drift, such as progressive degradation of static backgrounds, and (ii) high-level semantic drift, such as inconsistent character identity and view-dependent attributes. To address this issue, EverAnimate restores drifted flow trajectories by anchoring generation to a persistent latent context memory, consisting of two complementary mechanisms. (i) Persistent Latent Propagation maintains a context memory across chunks to propagate identity and motion in latent space while mitigating temporal forgetting. (ii) Restorative Flow Matching introduces an implicit restoration objective during sampling through velocity adjustment, improving within-chunk fidelity. With only lightweight LoRA tuning, EverAnimate outperforms state-of-the-art long-animation methods in both short- and long-horizon settings: at 10 seconds, it improves PSNR/SSIM by 8%/7% and reduces LPIPS/FID by 22%/11%; at 90 seconds, the gains increase to 15%/15% and 32%/27%, respectively.

Tool use extends large language models beyond parametric knowledge, but reliable execution requires balancing appropriate reasoning depth with strict structural validity. We approach this problem from a case-based perspective to present CAST, a case-driven framework that treats historical execution trajectories as structured cases. Instead of reusing raw exemplar outputs, CAST extracts case-derived signals to identify complexity profiles for estimating optimal reasoning strategies, alongside failure profiles to map likely structural breakdowns. The framework translates this knowledge into a fine-grained reward design and adaptive reasoning, enabling the model to autonomously internalize case-based strategies during reinforcement learning. Experiments on BFCLv2 and ToolBench demonstrate that CAST improves both schema-faithful execution and task-level tool-use success while reducing unnecessary deliberation. The approach achieves up to 5.85 percentage points gain in overall execution accuracy and reduces average reasoning length by 26%, significantly mitigating high-impact structural errors. Ultimately, this demonstrates how historical execution cases can provide reusable adaptation knowledge for calibrated tool use.

Agentic modeling aims to transform LLMs into autonomous agents capable of solving complex tasks through planning, reasoning, tool use, and multi-turn interaction with environments. Despite major investment, open research remains constrained by infrastructure and training gaps. Many high-performing systems rely on proprietary codebases, models, or services, while most open-source frameworks focus on orchestration and evaluation rather than scalable agent training. We present Orchard, an open-source framework for scalable agentic modeling. At its core is Orchard Env, a lightweight environment service providing reusable primitives for sandbox lifecycle management across task domains, agent harnesses, and pipeline stages. On top of Orchard Env, we build three agentic modeling recipes. Orchard-SWE targets coding agents. We distill 107K trajectories from MiniMax-M2.5 and Qwen3.5-397B, introduce credit-assignment SFT to learn from productive segments of unresolved trajectories, and apply Balanced Adaptive Rollout for RL. Starting from Qwen3-30B-A3B-Thinking, Orchard-SWE achieves 64.3% on SWE-bench Verified after SFT and 67.5% after SFT+RL, setting a new state of the art among open-source models of comparable size. Orchard-GUI trains a 4B vision-language computer-use agent using only 0.4K distilled trajectories and 2.2K open-ended tasks. It achieves 74.1%, 67.0%, and 64.0% success rates on WebVoyager, Online-Mind2Web, and DeepShop, respectively, making it the strongest open-source model while remaining competitive with proprietary systems. Orchard-Claw targets personal assistant agents. Trained with only 0.2K synthetic tasks, it achieves 59.6% pass@3 on Claw-Eval and 73.9% when paired with a stronger ZeroClaw harness. Collectively, these results show that a lightweight, open, harness-agnostic environment layer enables reusable agentic data, training recipes, and evaluations across domains.

We introduce TopoPrimer, a framework that makes the global topological structure of the series population an explicit input to any forecasting model. TopoPrimer improves accuracy across diverse domains, stabilizes forecasts under seasonal demand spikes, and closes the cold-start gap. Precomputed once per domain via persistent homology and spectral sheaf coordinates, TopoPrimer deploys per token for fully-trained models and as a lightweight adapter for pre-trained backbones. Of these two components, sheaf coordinates are the primary accuracy driver. Across four public benchmarks on Chronos and TimesFM, TopoPrimer consistently improves forecasting accuracy, with gains of up to 7.3% MSE on ECL. The topology advantage persists with near-identical magnitude across zero-shot and fine-tuned backbones, suggesting topology and per-series training capture complementary signals. The gains are most pronounced in difficult regimes. Under peak seasonal demand, classical and zero-shot models degrade by up to 50%, while TopoPrimer stays within 10%. At cold start with no item history, TopoPrimer reduces MAE by 27% over a topology-free baseline.

Large language models (LLMs) have been extensively studied from computational and cognitive perspectives, yet their behavior as communicative actors in socially structured contexts remains underexplored. This study examines whether LLM-based multi-agent systems exhibit systematic linguistic adaptation in response to perceived social observation contexts -- a question with direct implications for AI governance and auditing. Drawing on Habermas's (1981) Theory of Communicative Action, Goffman's (1959) dramaturgical model, Bell's (1984) Audience Design framework, and the Hawthorne Effect, we report a controlled experiment involving 100 multi-agent debate sessions across five conditions (n = 20 each). Conditions varied the framing of social observation -- from explicit monitoring by university researchers, to negation of monitoring, to an observer-substitution condition replacing human researchers with an automated AI auditing system. Monitored conditions (Delta+24.9%, Delta+24.2%) and the automated AI monitoring condition (Delta+22.2%) produce higher TTR change than audience-framing conditions (Delta+17.7%), F(4, 94) = 2.79, p = .031. Message length shows a fully dissociated effect, F(4, 95) = 19.55, p < .001. A fifth condition -- replacing human with AI observers -- yields intermediate TTR adaptation, suggesting LLM behavior is sensitive to observer identity: human evaluation elicits stronger register formalization than automated AI surveillance. We discuss implications for AI governance, algorithmic auditing, and the repositioning of LLMs as contextually sensitive communicative actors.

Intelligent Reflecting Surfaces (IRSs) are a promising technology for enhancing the spectral and energy efficiency of millimeter-wave (mmWave) multiple-input multiple-output (MIMO) systems. In these systems, accurate channel estimation remains challenging due to the passive nature of IRS elements and the high pilot overhead in large-scale deployments. This paper presents a deep learning-based Multi-Block Attention (MBA) framework for efficient cascaded channel estimation in IRS-assisted mmWave MIMO systems that utilize orthogonal frequency division multiplexing (OFDM). First, we show the optimality of the discrete Fourier transform (DFT) and Hadamard matrices as phase configurations for least squares (LS) estimation. To reduce training overhead, we selectively deactivate IRS elements and compensate for induced feature loss using a two-stage architecture: (i) a Convolutional Attention Network (CAN) for spatial correlation recovery and (ii) a Complex Multi-Convolutional Network (CMN) for noise suppression. The MBA architecture mitigates error propagation through attention-guided feature refinement and denoising. Simulation results indicate that the MBA method reduces pilot overhead by up to 87% compared to the LS estimator. Additionally, at signal-to-noise ratios of 10 dB, our proposed method achieves approximately 51% lower normalized mean squared error (NMSE) than leading methods. It also maintains low computational complexity and adapts effectively to various propagation environments.

Web agents can autonomously complete online tasks by interacting with websites, but their exposure to open web environments makes them vulnerable to prompt injection attacks embedded in HTML content or visual interfaces. Existing guard models still suffer from limited generalization to unseen domains and attack patterns, high false positive rates on benign content, reduced deployment efficiency due to added latency at each step, and vulnerability to adversarial attacks that evolve over time or directly target the guard itself. To address these limitations, we propose WARD (Web Agent Robust Defense against Prompt Injection), a practical guard model for secure and efficient web agents. WARD is built on WARD-Base, a large-scale dataset with around 177K samples collected from 719 high-traffic URLs and platforms, and WARD-PIG, a dedicated dataset designed for prompt injection attacks targeting the guard model. We further introduce A3T, an adaptive adversarial attack training framework that iteratively strengthens WARD through a memory-based attacker and guard co-evolution process. Extensive experiments show that WARD achieves nearly perfect recall on out-of-distribution benchmarks, maintains low false positive rates to preserve agent utility, remains robust against guard-targeted and adaptive attacks under substantial distribution shifts, and runs efficiently in parallel with the agent without introducing additional latency.

Online OS tuning can improve long-running services, but existing controllers are poorly matched to live hosts. They treat scheduler, power, memory, and I/O controls as black-box variables and optimize a scalar reward. This view ignores cross-knob policy structure, breaks down when application metrics are unavailable, and can send a running service into degraded regions that persist after the bad setting is removed. We present SemaTune, a host-side framework for steady-state OS tuning with bounded language-model guidance. SemaTune turns knob schemas, telemetry, current configuration, recent action--response history, and retrieved prior runs into a compact decision context. A fast loop proposes low-latency updates, a slower loop periodically revises the search strategy, and every proposed change passes through typed validation before reaching kernel or sysctl interfaces. This lets the controller reason about OS-control meaning and indirect performance signals while keeping model cost, latency, and authority constrained. We evaluate SemaTune on 13 live workloads from five benchmark suites while tuning up to 41 Linux parameters. Across the suite, SemaTune improves stable-phase performance by 72.5\% over default settings and by 153.3\% relative to the strongest non-LLM baseline. A 30-window session costs about \$0.20 in model calls. With only host-level metrics, SemaTune still outperforms baselines given direct application objectives by 93.7 percentage points, while avoiding severe degraded regions reached by structure-blind exploration.

Multimodal Retrieval-Augmented Generation (RAG) systems retrieve evidence at coarse granularities (entire images or scenes), creating a mismatch with fine-grained user queries and making failures unverifiable. We introduce GranuVistaVQA, a multimodal benchmark featuring real-world landmarks with element-level annotations across multiple viewpoints, capturing the partial observation challenge where individual images contain only subsets of entities. We further propose GranuRAG, a multi-granularity framework that treats visual elements as first-class retrieval units through three stages: element-level detection and classification, multi-granularity cross-modal alignment for evidence retrieval, and attribution-constrained generation. By grounding retrieval at the element level rather than relying on implicit attention, our approach enables transparent error diagnosis. Experiments demonstrate that GranuRAG achieves up to 29.2% improvement over six strong baselines for this task.

Shapley value and its priority-aware extensions are widely used for valuation in machine learning, but existing methods require pairwise priority to be binary and acyclic, a restriction spectacularly violated in real-data examples such as aggregated human preferences and multi-criterion comparisons. We introduce the generalized priority-aware Shapley value (GPASV), a random order value defined on arbitrary directed weighted priority graphs, in which pairwise edges penalize rather than forbid order violations. GPASV covers a range of classical models as boundary cases. We establish GPASV through an axiomatic characterization, develop the associated computational methods, and introduce a priority sweeping diagnostic extending PASV's. We apply GPASV to LLM ensemble valuation on the cyclic Chatbot Arena preference graph, illustrating that priority-aware valuation is not a one-button operation: different balances of pairwise graph priority versus individual soft priority produce substantively different valuations of the same data.

As large language models empower healthcare, intelligent clinical decision support has developed rapidly. Longitudinal electronic health records (EHR) provide essential temporal evidence for accurate clinical diagnosis and analysis. However, current large language models have critical flaws in longitudinal EHR reasoning. First, lacking fine-grained statistical reasoning, they often hallucinate clinical trends and metrics when quantitative evidence is textually implied, biasing diagnostic inference. Second, non-uniform time series and scarce labels in longitudinal EHR hinder models from capturing long-range temporal dependencies, limiting reliable clinical reasoning. To address the above limitations, this work presents the Probabilistic Chain-of-Thought Completion Agent (COTCAgent), a hierarchical reasoning framework for longitudinal electronic health records. It consists of three core modules. The Temporal-Statistics Adapter (TSA) converts analytical plans into executable code for standardized trend output. The Chain-of-Thought Completion (COTC) layer leverages a symptom-trend-disease knowledge base with weighted scoring to evaluate disease risk, while the bounded completion module acquires structured evidence through standardized inquiries and iterative scoring constraints to ensure rigorous reasoning. By decoupling statistical computation, feature matching, and language generation, the framework eliminates reliance on complex multi-modal inputs and enables efficient longitudinal record analysis with lower computational overhead. Experimental results show that COTCAgent powered by Baichuan-M2 achieves 90.47% Top-1 accuracy on the self-built dataset and 70.41% on HealthBench, outperforming existing medical agents and mainstream large language models. The code is available at https://github.com/FrankDengAI/COTCAgent/.

Generative AI increasingly supports educational design tasks, e.g., through Large Language Models (LLMs), demonstrating the capability to design assessment questions that are aligned with pedagogical frameworks (e.g., Bloom's taxonomy). However, they often rely on subjective or limited evaluation methods; focus primarily on proprietary models; or rarely systematically examine generation, evaluation, or deployment constraints in real educational settings. Meanwhile, Small Language Models (SLMs) have emerged as local alternatives that better address privacy and resource limitations; yet their effectiveness for assessment tasks remains underexplored. To address this gap, we systematically compare LLMs and SLMs for assessment question design; evaluate generation quality across Bloom's taxonomy levels using reproducible, pedagogically grounded metrics; and further assess model-based judging against expert-informed evaluation by analyzing reliability and agreement patterns. Results show that SLMs achieve competitive performance across key pedagogically motivated quality dimensions while enabling local, privacy-sensitive deployment. However, model-based evaluations also exhibit systematic inconsistencies and bias relative to expert ratings. These findings provide evidence to posit language models as bounded assistants in assessment workflows; underscore the necessity of Human-in-the-Loop; and advance the automated educational question generation field by examining quality, reliability, and deployment-aware trade-offs.

Reinforcement Learning with Verifiable Rewards (RLVR) has achieved great success in developing Large Language Models (LLMs) with chain-of-thought rollouts for many tasks such as math and coding. Nevertheless, RLVR struggles with sample efficiency on difficult problems where correct rollouts are hard to generate. Prior works propose to address this issue via demonstration-guided RLVR, i.e., to conduct Supervised FineTuning (SFT) when RL fails; however, SFT often requires a lot of data, which can be expensive to acquire. In this paper, we propose FEST, a FEw-ShoT demonstration-guided RLVR algorithm. It attains compelling results with only 128 demonstrations randomly selected from an SFT dataset. We find that three components are vital for the success: supervised signal, on-policy signal, and decaying weights on the few-shot SFT dataset to prevent overfitting from multiple-epoch training. On several benchmarks, FEST outperforms baselines with magnitudes less SFT data, even matching their performance with full dataset.

Scientific contributions rarely develop in isolation, but instead build upon prior discoveries. We formulate the task of automated technological roadmapping as extracting scientific contributions from scholarly articles and linking them to their prerequisites. We present the Scientific Contribution Graph, a large-scale AI/NLP-domain resource containing 2 million detailed scientific contributions extracted from 230k open-access papers and connected by 12.5 million prerequisite edges. We further introduce scientific prerequisite prediction, a scientific discovery task in which models predict which existing technologies can enable future discoveries, and show that contemporary models are rapidly improving on this task, reaching 0.48 MAP when evaluated using temporally filtered backtesting. We anticipate technological roadmapping resources such as this will support scientific impact assessment and automated scientific discovery.

The detection of Alzheimers disease (AD) is considered crucial, as timely intervention can improve patient outcomes. Electroencephalogram (EEG)-based diagnosis has been recognized as a non-invasive, accessible, and cost-effective approach for AD detection; however, it faces challenges related to data availability, accuracy of modern deep learning methods, and the time-consuming nature of expert-based interpretation. In this study, a novel lightweight and high-performance model, DeepTokenEEG, was designed for the diagnosis of AD and the classification of EEG signals from AD patients, individuals with other neurological conditions, and healthy subjects. Unlike traditional heavy-weight models, DeepTokenEEG ultilizes spatial and temporal tokenizer that effectively captures AD-related biomarkers in both temporal and frequency domain with only 0.29 million paramaters. Trained in a combined dataset of 274 subjects, including 180 AD cases, and 94 healthy controls, the proposed method achieves a maximum recorded accuracy of 100% on specific frequency bands, representing an improvement of 1.41-15.35% over state-of-the-art methods on the same dataset. These results indicate the potential of DeepTokenEEG for early detection and screening of AD, with promising applicability for deployment due to its compact size.

Premature closure, or committing to a conclusion before sufficient information is available, is a recognized contributor to diagnostic error but remains underexamined in large language models (LLMs). We define LLM premature closure as inappropriate commitment under uncertainty: providing an answer, recommendation, or clinical guidance when the safer response would be clarification, abstention, escalation, or refusal. We evaluated five frontier LLMs across structured and open-ended medical tasks. In MedQA (n = 500) and AfriMed-QA (n = 490) questions where the correct choice had been removed, models still selected an answer at high rates, with baseline false-action rates of 55-81% and 53-82%, respectively. In open-ended evaluation, models gave inappropriate answers on an average of 30% of 861 HealthBench questions and 78% of 191 physician-authored adversarial queries. Safety-oriented prompting reduced premature closure across models, but residual failure persisted, highlighting the need to evaluate whether medical LLMs know when not to answer.

As generative AI becomes increasingly integrated into journalism, designing effective AI-use disclosures that inform readers without imposing unnecessary burden is a key challenge. While prior research has primarily focused on trust and credibility, the impact of disclosures on readers' attentional and cognitive load remains underexplored. To address this gap, we conducted a $3\times2\times2$ mixed factorial study manipulating the level of AI-use disclosure detail (none, one-line, detailed), news type (politics, lifestyle), and role of AI (editing, partial content generation), measuring load via NASA-TLX and eye-tracking. Our results reveal a significant attentional cost: one-line disclosures resulted in significantly higher fixation durations and saccade counts, particularly for AI-edited content. Detailed disclosures did not impose additional burden. Drawing on Information-Gap Theory, we argue that brief labels may trigger increased visual scrutiny by alerting readers to AI use without providing enough information. NASA-TLX scores and pupil diameter showed no significant differences across conditions, suggesting that AI-use disclosures do not impose cognitive burden regardless of the detail level. Interview insights contextualize these findings and reveal a strong preference for detailed or ``detail-on-demand'' designs. Our findings inform the design of gaze-informed adaptive disclosure interfaces that dynamically adjust transparency levels based on readers' attentional patterns and news context.

From subcellular structures to entire organisms, many natural systems generate complex organisation through self-organisation: local interactions that collectively give rise to global structure without any blueprint of the outcome. Yet a significant portion of the information driving such processes is not produced by self-organisation itself, instead, it is often offloaded to initial conditions of the system. Biological development is a prime example, where maternal pre-patterns encode positional and symmetry-breaking information that scaffolds the self-organising process. From maternal morphogen gradients in early embryogenesis to tissue-level morphogenetic pre-patterns guiding organ formation, this transfer of information to initial conditions, analogous to a memory-compute trade-off in computational systems, is a fundamental part of developmental processes. In this work, we study this offloading phenomenon by introducing a model that jointly learns both the self-organisation rules and the pre-patterns, allowing their interplay to be varied and measured under controlled conditions: a Neural Cellular Automaton (NCA) paired with a learned coordinate-based pattern generator (SIREN), both trained simultaneously to generate a set of patterns. We provide information-theoretic analyses of how information is distributed between pre-patterns and the self-organising process, and show that jointly learning both components yields improvements in robustness, encoding capacity, and symmetry breaking over purely self-organising alternatives. Our analysis further suggests that effective pre-patterns do not simply approximate their targets; rather, they bias the developmental dynamics in ways that facilitate convergence, pointing to a non-trivial relationship between the structure of initial conditions and the dynamics of self-organisation.

Every day, users generate digital traces (e.g., social media posts, chats, and online interactions) that are inherently timestamped and may reflect aspects of their mental state. These traces can be organized into temporal trajectories that capture how a user's mental health signals evolve, including phases of improvement, deterioration, or stability. In this work, we propose an explainable framework for detecting and analyzing depression-related status shifts in user digital traces. The approach combines multiple BERT-based models to extract complementary signals across different dimensions (e.g., sentiment, emotion, and depression severity). Such signals are then aggregated over time to construct user-level trajectories that are analyzed to identify meaningful change points. To enhance interpretability, the framework integrates a large language model to generate concise and human-readable reports that describe the evolution of mental-health signals and highlight key transitions. We evaluate the framework on two social media datasets. Results show that the approach produces more coherent and informative summaries than direct LLM-based reporting, achieving higher coverage of user history, stronger temporal coherence, and improved sensitivity to change points. An ablation study confirms the contribution of each component, particularly temporal modeling and segmentation. Overall, the method provides an interpretable view of mental health signals over time, supporting research and decision making without aiming at clinical diagnosis.

Ovarian cancer is the most lethal gynecologic malignancy: around 60% of patients are diagnosed at an advanced stage, with an associated 5-year survival rate of about 30%. Early identification of non-responders to neoadjuvant chemotherapy remains a key unmet need, as it could prevent ineffective therapy and avoid delays in optimal surgical management. This work proposes a non-invasive deep learning framework to predict neoadjuvant chemotherapy response from pre-treatment contrast-enhanced CT by leveraging automatically derived 3D lesion masks. The approach encodes axial slices with a partially fine-tuned pretrained image encoder and aggregates slice-level representations into a volumetric embedding through an attention-based module. Training combines classification loss with supervised contrastive regularization and hard-negative mining to improve separation between ambiguous responders and non-responders. The method was developed on a retrospective single-center cohort from the European Institute of Oncology (Milan, IT), including 280 eligible patients (147 responder, 133 non-responder). On the test cohort, the model achieved a ROC-AUC of 0.73 (95% CI: 0.58-0.86) and an F1-score of 0.70 (95% CI: 0.56-0.82). Overall, these results suggest that the proposed architecture learns clinically relevant predictive patterns and provides a robust foundation for an imaging-based stratification tool.

Generating a street-level 3D scene from a single satellite image is a crucial yet challenging task. Current methods present a stark trade-off: geometry-colorization models achieve high geometric fidelity but are typically building-focused and lack semantic diversity. In contrast, proxy-based models use feed-forward image-to-3D frameworks to generate holistic scenes by jointly learning geometry and texture, a process that yields rich content but coarse and unstable geometry. We attribute these geometric failures to the extreme viewpoint gap and sparse, inconsistent supervision inherent in satellite-to-street data. We introduce Sat3DGen to address these fundamental challenges, which embodies a geometry-first methodology. This methodology enhances the feed-forward paradigm by integrating novel geometric constraints with a perspective-view training strategy, explicitly countering the primary sources of geometric error. This geometry-centric strategy yields a dramatic leap in both 3D accuracy and photorealism. For validation, we first constructed a new benchmark by pairing the VIGOR-OOD test set with high-resolution DSM data. On this benchmark, our method improves geometric RMSE from 6.76m to 5.20m. Crucially, this geometric leap also boosts photorealism, reducing the Fréchet Inception Distance (FID) from $\sim$40 to 19 against the leading method, Sat2Density++, despite using no extra tailored image-quality modules. We demonstrate the versatility of our high-quality 3D assets through diverse downstream applications, including semantic-map-to-3D synthesis, multi-camera video generation, large-scale meshing, and unsupervised single-image Digital Surface Model (DSM) estimation. The code has been released on https://github.com/qianmingduowan/Sat3DGen.

An index is a function that given an election outputs a value between 0 and 1, indicating the extent to which this election has a particular feature. We seek indices that capture agreement, diversity, and polarization among voters in approval elections, and that are normalized with respect to saturation. By the latter we mean that if two elections differ by the fraction of candidates approved by an average voter, but otherwise are of similar nature, then they should have similar index values. We propose several indices, analyze their properties, and use them to (a) derive a new map of approval elections, and (b) show similarities and differences between various real-life elections from Pabulib, Preflib and other sources.

We address the discounted reward setting in reinforcement learning (RL). To mitigate the value approximation challenges in policy gradient methods, actor-critic approaches have been developed and are known to converge to stationary points under suitable assumptions. However, these methods rely on first-order updates. In contrast, second-order optimization provides principled curvature-aware updates that are proven to accelerate convergence, but its application in RL is limited by the computational complexity of Hessian estimation. In this work, we analyze second-order approximations for the actor update that leverage the full curvature information of the objective as much as possible. A stable approximation requires treating the action-value function as locally constant with respect to policy parameters, which does not generally hold in policy gradient methods. We show that this approximation becomes well-justified under a two-timescale actor-critic framework, where the critic evolves on a faster timescale and can be treated as quasi-stationary during actor updates. Building on this insight, we formulate a second-order actor-critic method for the discounted reward setting that leverages Hessian-vector product (HVP) computations, resulting in a computationally efficient and stable second-order update.

Gromov--Wasserstein (GW) distances compare graphs, shapes, and point clouds through internal distances, without requiring a common coordinate system. This invariance is powerful, but discrete GW is a nonconvex quadratic optimal transport problem and is difficult to estimate at scale. We propose \emph{Distance-Matrix Wasserstein} (DMW), a hierarchy of Wasserstein statistics comparing laws of random finite distance matrices. Rather than optimizing a global point-level alignment, DMW samples $n$ points from each space, records their pairwise distances, and transports the resulting matrix laws. We prove that DMW is a relaxation and lower bound of GW, and establish a reverse approximation inequality: the GW--DMW gap is controlled by the Wasserstein error of approximating each original measure with $n$ samples. Hence population DMW converges to GW as sampled subspaces become dense. We further give finite-sample bounds, including intrinsic-dimensional rates that depend on the data manifold rather than the ambient matrix dimension $\binom n2$. For scalable computation, we introduce sliced and multi-scale DMW; for $p=1$, the sliced multi-scale dissimilarity yields positive-definite exponential kernels. Experiments on synthetic metric spaces, scalability benchmarks, graph classification, and two-sample testing validate the theory and demonstrate an interpretable GW-style proxy for structural comparison.

Analyzing microscopy images to extract biological object properties (e.g., their morphological organization, temporal dynamics, and population density) is fundamental to various biomedical research. Yet conducting this manually is costly and time-consuming. Though deep learning-based approaches have been explored to automate this process, the substantial diversity of microscopy analysis settings in practice (including variations of biological object types, sample processing protocols, imaging equipment, and analysis tasks, etc.) often renders them ineffective. As a result, these approaches typically require extensive adaptation for different settings, which, however, can impose burdens that are often practically unsustainable for laboratories, forcing biomedical researchers to still commonly rely on manual analysis, thereby severely bottlenecking the pace of biomedical research progress. This situation has created a pressing and long-standing need for a reliable and broadly applicable microscopy image analysis tool, yet such a tool is still missing. To address this gap, we present the first ready-to-use microscopy image analysis framework, MicroscopyMatching, that can reliably perform key analysis tasks (including segmentation, tracking, and counting) across diverse microscopy analysis settings. From a fundamentally different perspective, MicroscopyMatching reformulates diverse microscopy image analysis tasks as a unified matching problem, effectively handling this problem by exploiting the robust matching capability from pre-trained latent diffusion models.

Speculative decoding accelerates LLM inference by having a lightweight draft model propose speculative windows of candidate tokens for parallel verification by a larger target model. In practice, speculative efficiency is often bottlenecked by hard-to-draft positions, where an early mismatch truncates the accepted prefix and invalidates the rest of the speculative window. Most learning-based drafters are still optimized with token-level supervised objectives, even though speculative utility is inherently window-level and prefix-sensitive. We propose PPOW (Performance-Driven Policy Optimization with Adaptive Windowing), a reinforcement learning framework that shifts drafter optimization from token-level imitation to window-level optimization. PPOW combines a Cost-Aware Speedup Reward, a Distribution-Based Proximity Reward, and Adaptive Divergence-Aware Windowing, which prioritizes informative windows with high confidence-weighted draft-target divergence. PPOW achieves average acceptance lengths of 6.29-6.52 and speedups of 3.39-4.36$\times$ across multiple model families and benchmarks under a unified decoding protocol. These results show that performance-driven window-level optimization is a practical approach to improving speculative decoding efficiency.

A fundamental limitation of Text-to-Code is that no guarantee can be obtained about the correctness of the generated code. Therefore, to ensure its correctness, the generated code still has to be reviewed, tested, and maintained by developers. However, parsing through LLM-generated code can be tedious and time-consuming, potentially negating the productivity gains promised by AI-coding tools. To address this challenge, we present Viverra, a system that automatically produces formally verified annotations alongside generated code to aid user's understanding of the generated program. Given a natural-language task description, Viverra prompts an LLM to synthesize a C program together with candidate assertions expressing safety and correctness properties. It then verifies those assertions in a compositional and best-effort manner via a portfolio of bounded model checkers. Evaluation on 18 diverse programming tasks suggests that Viverra can efficiently generate code with verified assertions, and that these assertions improve users' performance on code-comprehension tasks in a user study with more than 400 participants.

GraphFlow is a visual workflow system designed to improve the reliability of agentic AI automation in multi-step, mission-critical processes. In these workflows, small errors compound rapidly: under an idealized model of independent steps, a ten-step process with 90% per-step reliability completes successfully only 35% of the time. Existing workflow platforms provide durable execution and observability but offer few semantic correctness guarantees, while agentic systems plan at inference time, making behavior sensitive to prompt variation and difficult to audit. GraphFlow is designed to address this gap by treating workflow diagrams as the executable specification, a single artifact defining data scope, execution semantics, and monitoring. At compile time, a restricted class of diagrams is specified to produce reusable automations whose contracts (preconditions, postconditions, and composition obligations) are intended to be proof-checked before admission to a shared library. At runtime, a durable engine records outcomes in an append-only event log and can enforce contracts at system boundaries, supporting replay, retries, and audit. Swimlanes make trust boundaries explicit, separating verified logic from external systems, human judgment, and AI decisions. A year-long pilot across three clinical sites executed 8,728 cohort-enrolled workflow runs with a 97.08% completion rate under an early prototype without the verified-core subsystem; observed failures were localized primarily to external integrations. The formal semantics and proof-checked admission model described here are specified and under active development. Evaluation of the verified core is reserved for future work.

Supervised fine-tuning (SFT) provides the standard approach for teaching LLMs new behaviors from offline expert demonstrations. However, standard SFT uniformly fits all samples -- including those with low likelihood under the base model -- which can disproportionately drive training updates toward overfitting specific samples rather than learning the target behavior. Moreover, adapting to these unlikely samples induces substantial policy shifts that degrade prior capabilities. Existing methods mitigate this by filtering, regenerating, or down-weighting low-likelihood data. In doing so, they often suppress precisely the novel behaviors the base model has yet to learn. We propose InfoSFT, a principled weighting scheme for the SFT objective that concentrates learning signals on maximally informative, medium-confidence tokens -- those neither overly familiar to the base model nor too unlikely to cause instability. Requiring only a one-line modification to the standard token-wise loss, InfoSFT demonstrably improves generalization over vanilla SFT and likelihood-weighted baselines across math, code, and chain-of-thought tasks with diverse model families, while better preserving pre-existing capabilities.

Large vision-language models (LVLMs) have achieved remarkable performance across diverse multimodal tasks, yet they continue to suffer from hallucinations, generating content that is inconsistent with the visual input. Prior work DHCP (Detecting Hallucinations by Cross-modal Attention Pattern) has explored hallucination detection from the perspective of cross-modal attention, but does not address hallucination mitigation. In this paper, we propose MHSA (Mitigating Hallucinations via Steered Attention), a lightweight framework that mitigates hallucinations by learning to correct cross-modal attention patterns in LVLMs. MHSA trains a simple three-layer MLP generator to produce corrected attention, guided by supervisory signals from the DHCP discriminator and the LVLM itself. During inference, MHSA mitigates both discriminative and generative hallucinations across various datasets and LVLMs by simply replacing the original cross-modal attention with the corrected one, without modifying any LVLM parameters. By extending cross-modal attention mechanisms from hallucination detection to hallucination mitigation, MHSA offers a novel perspective on hallucination research in LVLMs and helps enhance their reliability.

We address the problem of conformal selection, where an agent must select a minimal subset of options to ensure that at least one ``success'' is identified with a pre-specified target probability $φ$. While traditional online conformal prediction focuses on maintaining validity for the observed sequence, minimizing the resource cost (efficiency) of such selections, especially under limited feedback, remains a significant challenge. In this work, we consider settings with the most limited ``bandit'' feedback, and demonstrate that the simple Adaptive Conformal Inference (ACI) update rule, when applied to the appropriate control parameter or dual variable, is both adversarially valid, ensuring the success target is met on average for any input sequence (and hence under distribution shifts), and stochastically efficient, achieving sublinear efficiency regret for $i.i.d.$ inputs against an appropriate stochastic benchmark. We show such guarantees under canonical models capturing bandit and semi-bandit feedback to the agent via a unifying algorithmic technique, and analytic framework involving Lyapunov functions. Our approach handles more complex settings than prior work, while requiring significantly less feedback, and our results provide a new theoretical bridge between efficient online learning with limited feedback and distribution-free uncertainty quantification.

Electroencephalogram (EEG) signals are highly susceptible to artifacts, resulting in a low signal-to-noise ratio which makes extraction of meaningful neural information challenging. Artifact Subspace Reconstruction (ASR) is one of the most widely used artifact filtering techniques in EEG-based BCI applications, owing to its real-time applicability. ASR reconstructs artifact-free signals by operating in Principal Component (PC) space within sliding windows. However, ASR performance is critically sensitive to its threshold parameter - an incorrect threshold risks removing task-relevant neural features alongside artifacts. Furthermore, since PCs are linear combinations of all channels, subspace reconstruction in PC space may alter the underlying data structure, potentially discarding essential neural information. To address these limitations, we propose nASR, a novel end-to-end trainable Keras layer that jointly optimizes artifact rejection and downstream decoding. nASR introduces two trainable threshold parameters: K, which governs artifact detection in PC variance space, and L, which quantifies eigen-spread to pinpoint the primary artifact--contributing channels, enabling selective channel-level reconstruction that preserves clean channel information. An ablation study comprising five model variants (m01 - m05), evaluated across two subjects from the BCI Competition IV Dataset 1, confirms that nASR variants consistently outperform traditional ASR on test classification metrics, while achieving a 6-8x reduction in inference time, making nASR a strong candidate for real-time BCI applications demanding both low latency and high decoding performance.

Semantic communication systems for goal-oriented transmission must protect task-relevant information not only through source compression but also via physical layer mapping. Existing approaches decouple constellation design and semantic encoding, exposing critical symbols to channel errors at the same rate as irrelevant ones. Contrary to this, in this paper, a joint semantic-physical layer framework is proposed, which is composed of a vector quantized-variational autoencoder that extracts discrete latent concepts, a semantic criticality indicator (SCI) that scores each concept by task relevance, and a deep reinforcement learning agent that dynamically selects the transmission subset based on instantaneous channel conditions. At the physical layer, a learned semantic-aware M -QAM constellation assigns symbol positions according to joint co-occurrence statistics and SCI scores, departing from the uniform spacing and Gray coding of standard M -QAM which minimizes average BER without regard for semantic content. We introduce a novel semantic symbol vulnerability (SSV) metric and a semantic protection probability (SPP) to quantify the exposure of task-critical symbols to decoding errors, and prove that any Gray-coded constellation is strictly suboptimal in SCI-Weighted SSV whenever the source exhibits non-uniform semantic importance and co-occurrence statistics. Simulation results demonstrate that the proposed constellation achieves near 100% SPP across modulation orders from 4-QAM to 1024-QAM versus 50% for standard constellations at high spectral efficiency, a 21:1 compression ratio with semantic quality above 0.9, generalizing across MNIST, Fashion-MNIST, and FSDD without modification.

Reliable position and shape control in tokamak plasmas requires accurate real-time regulation of several strongly coupled shape parameters. The control vectors that disentangle these couplings, referred to as \textit{virtual circuits} (VCs), enable independent shape parameter control for a specific Grad--Shafranov (GS) equilibrium. Numerical calculation of VCs is not currently feasible in real time, therefore VCs are usually computed prior to each experiment, using a small number of reference GS equilibria sampled along the desired scenario trajectory, with each VC used to control the plasma within a preset time interval. While effective near the reference equilibrium, this approach can lead to degraded performance as the plasma departs from the reference equilibrium and/or from the desired trajectory, and it complicates the design of robust control strategies for rapidly evolving plasma configurations. In this paper, we construct neural-network-based emulators of plasma shape parameters from which VCs can be derived, to provide the MAST Upgrade (MAST-U) plasma control system with state-aware VCs in real-time. To do this, we develop an extensive library of over a million simulated GS equilibria, covering a substantial portion of the MAST-U operational space. These emulators provide differentiable functions whose gradients can be rapidly computed, enabling the derivation of accurate VCs for real-time shape control. We perform extensive verification of the emulated VCs by testing whether they disentangle the control problem. The neural-network-based approach delivers high accuracy and orthogonality across a diverse range of equilibria. This work establishes the physical validity of emulated VCs as a scalable and general alternative to schedules of precomputed VCs.

Continual learning in multimodal large language models (MLLMs) aims to sequentially acquire knowledge while mitigating catastrophic forgetting, yet existing methods face inherent limitations: architecture-based approaches incur additional computational overhead and often generalize poorly to new tasks, rehearsal-based methods rely on storing historical data, raising privacy and storage concerns, and conventional regularization-based strategies alone are insufficient to fully prevent parameter interference. We propose Octopus, a two-stage continual learning framework based on History-Free Gradient Orthogonalization (HiFGO), which enforces gradient-level orthogonality without historical task data. Our proposed two-stage finetuning strategy decouples task adaptation from regularization, achieving a principled balance between plasticity and stability. Experiments on UCIT show that Octopus establishes state-of-the-art performance, surpassing prior SOTA by 2.14% and 6.82% in terms of Avg and Last.

Predictive world models enable agents to model scene dynamics and reason about the consequences of their actions. Inspired by human perception, object-centric world models capture scene dynamics using object-level representations, which can be used for downstream applications such as action planning. However, most object-centric world models and reinforcement learning (RL) approaches learn reactive policies that are fixed at inference time, limiting generalization to novel situations. We propose Slot-MPC, an object-centric world modeling framework that enables planning through Model Predictive Control (MPC). Slot-MPC leverages vision encoders to learn slot-based representations, which encode individual objects in the scene, and uses these structured representations to learn an action-conditioned object-centric dynamics model. At inference time, the learned dynamics model enables action planning via MPC, allowing agents to adapt to previously unseen situations. Since the learned world model is differentiable, we can use gradient-based MPC to directly optimize actions, which is computationally more efficient than relying on gradient-free, sampling-based MPC methods. Experiments on simulated robotic manipulation tasks show that Slot-MPC improves both task performance and planning efficiency compared to non-object-centric world model baselines. In the considered offline setting with limited state-action coverage, we find that gradient-based MPC performs better than gradient-free, sampling-based MPC. Our results demonstrate that explicitly structured, object-centric representations provide a strong inductive bias for controllable and generalizable decision-making. Code and additional results are available at https://slot-mpc.github.io.

Scaled Outer Product (SOP) is a post-training quantization methodology for large language model weights, designed to deliver near-lossless fidelity at 4.5--6 bits per weight on hardware with per-layer LUT decode. The methodology combines per-layer search of fixed and dynamic codebook pairs selected by a per-block selection bit, signed per-block scales, activation-weighted cosine selection, and multiple-choice knapsack promotion of sensitive layers with outlier and sparse-residual correction. Fixed codebooks include NF4, BOF4, Split87, and SH4; per-layer optimized codebooks (DD4) are hosted in LUT SRAM. A new hardware-efficient LUT output format (HIF) is proposed to improve performance, energy, and cost. Across six open model families, the recommended FP6 operating point (E2M3sUE4M4, 6.5 bpw) achieves lower weight reconstruction error than the conventional per-layer-POT FP8 baseline (E4M3, 8.0 bpw) at 1.5 bpw lower storage cost, demonstrating that block-scaled small atoms with carefully chosen scale precision can replace conventionally-deployed FP8. Full evaluation across the 4.5--6 bpw range, including layer promotion and sparse residual correction, is reported in a companion paper.

Recent advances in vision-language models (VLMs) have achieved impressive results on standard image-text tasks, yet their potential for visual procedure question answering (VP-QA) remains largely unexplored. VP-QA presents unique challenges where users query next-step actions by uploading images for intermediate states of complex procedures. To systematically evaluate VLMs on this practical task, we propose ProcedureVQA, a novel multimodal benchmark specifically designed for visual procedural reasoning. Through comprehensive analysis, we identify two critical limitations in current VLMs: inadequate cross-modal retrieval of structured procedures given visual states, and misalignment between image sequence granularity and textual step decomposition. To address these issues, we present Chain-of-Procedure (CoP), a hierarchical reasoning framework that first retrieves relevant instructions using visual cues, then performs step refinement through semantic decomposition, and finally generates the next step. Experiments across six VLMs demonstrate CoP's effectiveness, achieving up to 13% absolute improvement over standard baselines.

The success of deep learning in high-dimensional settings is often attributed to the presence of low-dimensional structure in real-world data. While standard theoretical models typically assume that this structure lies in the target function, projecting unstructured inputs onto a low-dimensional subspace, data such as images, text or genomic sequences exhibit strong spatial correlations within the input space itself. In this paper, we propose a tractable model to study how these correlations affect the sample complexity of learning with gradient descent on shallow neural networks. Specifically, we consider targets that depend on a small number of latent Boolean variables, and input features grouped into clusters and correlated with the latent variables. Under an identifiability assumption, we show that for a layerwise gradient-descent variant, the sample complexity scales with the number of hidden variables and, when the signal-to-noise ratio is sufficiently high, is independent of the input dimension, up to logarithmic terms. We empirically test our theoretical findings on both synthetic and real data.

Drone-view geo-localization aims to match a query drone image, often captured under adverse weather conditions (e.g., rain, snow, fog), against a gallery of geo-tagged satellite images. Weather-induced degradations in the drone view, such as noise, reduced visibility, and partial occlusions, severely exacerbate the intrinsic cross-view domain gap. While prior methods predominantly rely on weather-specific architectures or data augmentations, they have largely overlooked road map data, a readily available modality that provides strong, inherently weather-invariant geometric layout cues (e.g., road networks and building footprints) at negligible additional cost. We introduce GeoFuse, a cross-modal fusion framework that integrates precisely aligned road map tiles with satellite imagery to yield more discriminative and weather-resilient representations. We first augment the existing University-1652 and DenseUAV benchmarks with geo-aligned road maps, supplying structural priors robust to meteorological variations. Building on this, we propose a flexible fusion module that combines satellite and road map features via token-level and channel-level interactions, with a lightweight dynamic gating mechanism that adaptively weights modality contributions per instance. Finally, we employ class-level cross-view contrastive learning to promote robust alignment between weather-degraded drone features and the fused satellite-roadmap representations. Extensive experiments under diverse weather conditions show that GeoFuse consistently outperforms state-of-the-art methods, achieving +3.46% and +23.18% Recall@1 accuracy on the University-1652 and DenseUAV benchmarks, respectively.

Active learning for continuous regression has lacked an acquisition function that targets epistemic uncertainty when the predictive distribution is multimodal: variance misses modal disagreement, and information-theoretic targets like BALD are designed for discrete outputs. We introduce a Two-Index framework that makes this separation explicit: one stochastic index selects among competing model hypotheses (epistemic source), while a second governs within-hypothesis randomness (aleatoric source). An entropy decomposition within the framework identifies the mutual information between the output and the epistemic index as a principled acquisition objective, and we prove this quantity vanishes as the model is trained on growing datasets, confirming that it captures exactly the uncertainty data can resolve. Because this mutual information is intractable for continuous outputs, we derive the Mutual Information Lower Bound (MI-LB) acquisition function, a closed-form approximation for Mixture Density Network ensembles. On benchmarks featuring multimodal systems, MI-LB matches or beats every baseline evaluated and is the only method to do so consistently -- geometric and Fisher-based baselines compete only when the input space already encodes the multimodality, and collapse otherwise.

Imbalanced learning remains a fundamental challenge in tabular data applications. Despite decades of research and numerous proposed algorithms, a systematic empirical understanding of how different imbalanced learning methods behave across diverse data characteristics is still lacking. In particular, it remains unclear how different method families compare in predictive performance, robustness under varying data characteristics, and computational scalability. In this work, we present Tabular Imbalanced Learning Benchmark (TILBench), a large-scale empirical benchmark for tabular imbalanced learning. TILBench evaluates more than 40 representative algorithms across 57 diverse tabular datasets, resulting in over 200000 controlled experiments across a wide range of data characteristics. Our findings show that no single method consistently dominates across all settings; instead, the effectiveness of imbalanced learning methods depends strongly on dataset characteristics and computational constraints. Based on these findings, we provide practical recommendations for selecting appropriate methods in real-world applications.

Pluralistic alignment is typically operationalised as preference aggregation: producing responses that span (Overton), steer toward (Steerable), or proportionally represent (Distributional) diverse human values. We argue that aggregation alone is an incomplete primitive for deployed pluralistic alignment. Under genuine value pluralism, the failure mode of contemporary RLHF-trained assistants is not insufficient coverage but sycophantic consensus: a learned tendency to agree with, validate, and minimise friction with the immediate interlocutor. Because deployed AI systems now mediate consequential deliberation across health, civic life, labour, and governance, the collapse of disagreement at the interaction layer is not a narrow technical concern but a structural failure with distributive consequences. We reframe pluralistic alignment around three conversational mechanisms drawn from Grice's maxims: scoping (acknowledging the limits of one's perspective), signalling (surfacing value-conflict rather than smoothing it over), and repair (revising one's position on principled grounds, not on user pressure). We formalise a metric, the Pluralistic Repair Score (PRS), distinguishing principled revision from capitulation, and present a small-scale empirical illustration on two frontier RLHF-trained models (Claude Sonnet 4.5, N=198; GPT-4o, N=100) showing that, for both, agreement-following coexists with low repair-quality on contested-value prompts. PRS measures an interactional precondition for pluralism (visible disagreement; principled revision) rather than pluralism in full; we discuss the difference, take seriously the reflexive question of whose "principled" counts, and argue that pluralism is most decisively made or unmade at the deployment-governance layer: interfaces, preference-data pipelines, and audit infrastructure.

Knowledge graph (KG) foundation models aim to generalize across graphs with unseen entities and relations by learning transferable relational structure. However, most existing methods primarily emphasize relation-level universality, while in-context learning, the other pillar of foundation models remains under-explored for KG reasoning. In KGs, context is inherently structured and heterogeneous: effective prediction requires conditioning on the local context around the query entities as well as the global context that summarizes how a relation behaves across many instances. We propose KGPFN, a KG foundation model using Prior-data Fitted Network that unifies transferable relational regularities with inference-time in-context learning from structured context. KGPFN first learns relation representations via message passing on relation graphs to capture cross-graph relational invariances. For query-specific reasoning, it encodes local neighborhoods using a multi-layer NBFNet as local context. To enable ICL at global scale, it constructs relation-specific global context by retrieving a large set of instances of the query relation together with their local neighborhoods, and aggregates them within a Prior-Data Fitted Network framework that combines feature-level and sample-level attention. Through multi-graph pretraining on diverse KGs, KGPFN learns when to instantiate reusable patterns and when to override them using contextual evidence. Experiments on 57 KG benchmarks demonstrate that KGPFN achieves strong adaptation to previously unseen graphs through in-context learning alone, consistently outperforming competitive fine-tuned KG foundation models. Our code is available at https://github.com/HKUST-KnowComp/KGPFN.

Knowledge Graphs (KGs) are extensively used across different domains and in several applications. Often, these KGs are very large in size. Such KGs become unwieldy for tasks such as question answering and visualization. Summarization of KGs offers a viable alternative in such cases. Furthermore, personalized KG summarization is crucial in the current data-driven world as it captures the specific requirements of users based on their query patterns. Since it only maintains relevant information, the personalized summaries of KG are small, resulting in significantly smaller storage requirements and query runtime. In this work, we adapt the coreset theory to create personalized KG summaries. For a given dataset and a user-specific query workload, we present an approach that samples a relevant subset of triples using sensitivity-based importance sampling. We ensure that the subset approximates the characteristics of the full dataset with bounded approximation error. We define sensitivity scores that measure the importance of a triple with respect to a user's query workload, which are then used by our coreset construction algorithm. We explicitly focus on personalized knowledge graph summarization by constructing summaries independently for each user based on their query behaviour. Our evaluation on Freebase, WikiData, and DBpedia shows that COREKG delivers higher query-answering accuracy and structural coverage than the state-of-the-art methods, such as GLIMPSE, PPR, iSummary, PEGASUS and APEX$^2$ while requiring only a tiny fraction of the original graph.

Despite many successful attempts at explaining Deep Reinforcement Learning policies using distillation, it remains difficult to balance the performance-interpretability trade-off and select a fitting surrogate model. In addition to this, traditional distillation only minimizes the distance between the behavior of the original and the surrogate policy while other RL-specific components such as action value are disregarded. To solve this, we introduce a new model-agnostic method called Critic-Driven Voronoi State Partitioning, which partitions a black box control policy into regions where a simple class of model can be optimized using gradient descent. By exploiting the critic value network of the original policy, we iteratively introduce new subpolicies in regions with insufficient value, standing in for a measure of policy complexity. The partitioning, a Voronoi quantizer, uses nearest neighbor lookups to assign a linear function to each point in the state space resulting in a cell-like diagram. We validate our approach on several well known benchmarks and proof that this distillation approaches the original policy using a reasonable sized set of linear functions.

This paper presents a system for the 2024 Text-Dependent Speaker Verification (TdSV) Challenge. The system achieved a Minimum Detection Cost Function (MinDCF) of 0.0461 and an Equal Error Rate (EER) of 1.3\%. Our approach focused on adapting existing state-of-the-art neural networks, ResNet-TDNN and NeXt-TDNN, originally trained on the VoxCeleb dataset. This strategy was chosen because of the limited challenge duration and the available resources at the time. In addition, we designed a lightweight and resource-efficient model, EfficientNet-A0, trained specifically on the challenge dataset to improve adaptation and strengthen the ensemble approach. Our system combines advanced neural architectures, extensive data augmentation, and optimised hyperparameters. These components helped achieve strong performance in text-dependent speaker verification. The results also demonstrate the effectiveness of multi-model ensemble learning for both speaker and phrase verification.

Contrastively pre-trained Vision-Language Models (VLMs) serve as powerful feature extractors. Yet, their shared latent spaces are prone to structural anomalies and act as repositories for non-semantic, multi-modal noise. To address this phenomenon, we employ spectral decomposition of covariance matrices to decompose the VLM latent space into a multi-modal semantic signal component and a shared noise subspace. We observe that this noise geometry exhibits strong subgroup invariance across distinct data subsets. Crucially, pruning these shared noise dimensions is mainly harmless, preserving or actively improving downstream task performance. By isolating true semantic signals from artifactual noise, this work provides new mechanistic insights into the representational structure of modern VLMs, suggesting that a substantial fraction of their latent geometry is governed by shared, architecture-level noise rather than task-relevant semantics alone.

LLM-based autonomous agents have demonstrated strong capabilities in reasoning, planning, and tool use, yet remain limited when tasks require sustained coordination across roles, tools, and environments. Multi-agent systems address this through structured collaboration among specialized agents, but tighter coordination also amplifies a less explored risk: errors can propagate across agents and interaction rounds, producing failures that are difficult to diagnose and rarely translate into structural self-improvement. Existing surveys cover individual agent capabilities, multi-agent collaboration, or agent self-evolution separately, leaving the causal dependencies among them unexamined. This survey provides a unified review organized around four causally linked stages, which we term the LIFE progression: Lay the capability foundation, Integrate agents through collaboration, Find faults through attribution, and Evolve through autonomous self-improvement. For each stage, we provide systematic taxonomies and formally characterize the dependencies between adjacent stages, revealing how each stage both depends on and constrains the next. Beyond synthesizing existing work, we identify open challenges at stage boundaries and propose a cross-stage research agenda for closed-loop multi-agent systems capable of continuously diagnosing failures, reorganizing structures, and refining agent behaviors, extending current coordination frameworks toward more self-organizing forms of collective intelligence. By bridging these previously fragmented research threads, this survey aims to offer both a systematic reference and a conceptual roadmap toward autonomous, self-improving multi-agent intelligence.

Foundation models tokenize Ukrainian legal text with vastly different efficiency, yet no systematic comparison exists for this domain. We benchmark seven models from five providers on 273 validated court decisions from Ukraine's state registry (EDRSR), measuring tokenizer fertility and zero-shot performance on three tasks. Three findings emerge. (1) Tokenizer fertility varies 1.6x: Qwen3 models consume 60% more tokens than Llama-family models on identical input, directly reducing API cost. (2) NVIDIA Nemotron Super 3 (120B) achieves the highest composite score (83.1), outperforming Mistral Large 3 (675B total, 41B active) -- a model with 5.6x more total parameters and 3.4x more active parameters per token -- at one-third the API cost. (3) Few-shot prompting degrades performance by up to 26 percentage points; stratified and prompt-sensitivity ablations confirm this is intrinsic to Ukrainian-language demonstrations, not an artifact of example selection. For practitioners: tokenizer analysis should precede model selection, and zero-shot is a more reliable default than few-shot for morphologically rich languages.

Online surgical phase recognition (SPR) underpins context-aware operating-room systems and requires committing to a prediction at every frame from past context alone. Surgical video poses three demands that natural-video recognizers do not jointly address: procedures span tens of thousands of frames, time flows non-uniformly as long routine stretches are punctuated by brief phase-defining transitions, and the visual domain is narrow so backbone features are strongly correlated across channels. Existing recognizers either let per-frame cost grow with elapsed length, or hold cost bounded but advance state at a uniform rate with channel-independent dynamics, leaving the latter two demands unaddressed. We present SurgicalMamba, a causal SPR model built on Mamba2's structured state-space duality (SSD) that holds per-frame cost at O(d). It introduces three SSD-compatible components, each targeting one demand: a dual-path SSD block that separates long- and short-term regimes at the level of recurrent state; intensity-modulated stepping, a continuous-time time-warp that adapts the slow path's effective rate to phase-relevant information; and state regramming, a per-chunk Cayley rotation that opens cross-channel mixing in the otherwise axis-aligned SSM recurrence. The learned rotation planes inherit a phase-aligned structure without any direct supervision, offering an interpretable internal signature of surgical workflow. Across seven public SPR benchmarks, SurgicalMamba reaches state-of-the-art accuracy and phase-level Jaccard under strict online evaluation: 94.6%/82.7% on Cholec80 (+0.7 pp/+2.2 pp over the strongest prior) and 89.5%/68.9% on AutoLaparo (+1.7 pp/+2.0 pp), at 119 fps on a single GPU. Ablations isolate the contribution of each component. The code is publicly available at https://github.com/sukjuoh/Surgical-Mamba.

Speech-based analysis offers a scalable and non-invasive approach for detecting cognitive decline, yet progress has been constrained by the limited availability of clinically validated datasets collected under realistic conditions. We introduce PROCESS-2, a large-scale speech dataset designed to support research on automatic assessment of cognitive impairment from spontaneous and task-oriented speech. The dataset comprises recordings from 200 healthy controls, 150 mild cognitive impairment, and 50 dementia diagnoses collected using the CognoMemory digital assessment platform. Each participant completed a single assessment session, including picture description and verbal fluency tasks, accompanied by manually verified transcripts and participant-level metadata. PROCESS-2 contains approximately 21 hours of speech audio with predefined train/test partitions. Comprehensive technical validation evaluated demographic balance, clinical consistency, recording stability, embedding-space structure, and reproducible baseline modelling performance, demonstrating clinically meaningful group separation and stable performance across modelling approaches while preserving real-world conversational variability. PROCESS-2 is released under controlled access via Hugging Face to enable responsible reuse while protecting participant privacy, providing a reproducible benchmark resource for speech-based cognitive assessment research.

Electrocardiogram (ECG) monitoring in Internet of Medical Things (IoMT) networks is constrained by strict data-sharing regulations and privacy concerns. Federated learning (FL) enables collaborative learning by keeping raw ECG data on devices, but frequent transmissions of high-dimensional model updates incur heavy per-round traffic over bandwidth-limited links. To alleviate this bottleneck, federated distillation (FD) replaces parameter exchange with logit-based knowledge transfer. However, the performance of FD often degrades under the non-independent and identically distributed (non-IID) and long-tailed label distributions in ECG deployments. To address these challenges, we propose a bidirectional federated knowledge distillation (BiFedKD) framework that employs an aggregation-by-distillation pipeline with temperature scaling to produce a stable global distillation signal for cross-client alignment. Experiments on the MIT-BIH Arrhythmia dataset show that BiFedKD improves accuracy and Macro-F1 over the baseline by $3.52\%$ and $9.93\%$, respectively. Moreover, to reach the same Macro-F1, BiFedKD reduces communication overhead by $40\%$ and computation cost by $71.7\%$ compared with the baseline.

Graph Neural Networks (GNNs) have advanced significantly in handling graph-structured data, but a comprehensive framework for evaluating explainability remains lacking. Existing evaluation frameworks primarily involve post-hoc explanations, and operate in the setting where multiple methods generate a suite of explanations for a single model. This makes comparison of explanations across models difficult. Evaluation of inherently interpretable models often targets a specific aspect of interpretability relevant to the model, but remains underdeveloped in terms of generating insight across a suite of measures. We introduce AIM, a comprehensive framework that addresses these limitations by measuring Accuracy, Instance-level explanations, and Model-level explanations. AIM is formulated with minimal constraints to enhance flexibility and facilitate broad applicability. Here, we use AIM in a pipeline, extracting explanations from inherently interpretable GNNs such as graph kernel networks (GKNs) and prototype networks (PNs), evaluating these explanations with AIM, identifying their limitations and obtaining insights to their characteristics. Taking GKNs as a case study, we show how the insights obtained from AIM can be used to develop an updated model, xGKN, that maintains high accuracy while demonstrating improved explainability. Our approach aims to advance the field of Explainable AI (XAI) for GNNs, providing more robust and practical solutions for understanding and improving complex models.

Mild traumatic brain injury (mTBI) is a prevalent condition that remains difficult to diagnose in its early stages. Oculomotor dysfunction is a well-established marker of mTBI, motivating the development of portable tools that capture both eye-movement behavior and underlying neurophysiology. In this work, we present an initial framework that integrates electroencephalogram (EEG) with augmented-reality (AR)-based Vestibular/Ocular Motor Screening (VOMS) tasks to estimate subject-specific ocular response times. Pre-processed EEG signals, obtained through band-pass filtering and average referencing, are analyzed using a Redundant Discrete Wavelet Transform (RDWT)-driven deep neural framework. The RDWT coefficients are subjected to trainable zero-phase convolutional filtering and reconstructed into the time domain via inverse RDWT, followed by channel-wise temporal and spatial filtering using 2D convolution layers and convolutional-LSTM-based decoding. An ablation study demonstrates that wavelet-domain filtering serves as an effective denoising strategy, improving prediction performance. Sliding-window predictions were validated using Pearson correlation (>= 0.5), and Dynamic Time Warping (DTW) was subsequently used to estimate ocular response times. DTW-derived metrics revealed significant inter-subject differences across all VOM tasks, supported by Mann-Whitney U tests. Cross-correlation analysis further revealed task-dependent temporal behaviors: pursuit tasks exhibited reactive tracking, whereas saccades showed anticipatory responses. Overall, the results highlight pursuit tasks as particularly informative for distinguishing timing differences and demonstrate the potential of RDWT-based EEG features combined with DTW metrics for multimodal mTBI assessment.

Recent advances in 3D Gaussian Splatting (3DGS) have achieved remarkable success in high-fidelity Novel View Synthesis (NVS), yet the optimization process inevitably introduces noisy Gaussian primitives due to the sparse and incomplete initialization from Structure-from-Motion (SfM) point clouds. Most existing methods focus solely on adjusting the positions of primitives during optimization, while neglecting the underlying spatial structure. To this end, we introduce a new perspective by formulating the optimization of 3DGS as a primitive denoising process and propose Denoising-GS, a spatial-aware denoising framework for Gaussian primitives by taking both the positions and spatial structure into consideration. Specifically, we design an optimizer that preserves the spatial optimization flow of primitives, facilitating coherent and directed denoising rather than random perturbations. Building upon this, the Spatial Gradient-based Denoising strategy jointly considers the spatial supports of primitives to ensure gradient-consistent updates. Furthermore, the Uncertainty-based Denoising module estimates primitive-wise uncertainty to prune redundant or noisy primitives, while the Spatial Coherence Refinement strategy selectively splits primitives in sparse regions to maintain structural completeness. Experiments conducted on three benchmark datasets demonstrate that Denoising-GS consistently enhances NVS fidelity while maintaining representation compactness, achieving state-of-the-art performance across all benchmarks. Source code and models will be made publicly available.

A plethora real-world environments require agents to compete repeatedly for the same limited resource, calling for a temporal notion of fairness judged across entire interaction histories. This paper advances the theory of temporal fair division by introducing Rotational Periodicity (RP), a family of lightweight metrics, alongside the ALT family of sliding-window measures, within a unified framework for repeated multi-agent resource competition. We formalise the Multi-Agent Battle of the Exes (MBoE) as a repeated fair division instance and establish Perfect Alternation (PA) as its canonical temporally fair solution, drawing connections to proportionality, envy-freeness, and n-periodic round-robin allocation. RP decomposes temporal fairness into two complementary sub-measures: Rotational Score (RS) and Waiting Periods Evaluation (WPE), achieving O(nu+n) time complexity versus the O(nu*n) of ALT, where nu is the episode count and n the agent count. Empirical evaluation across n in {2,3,5,8,10} reveals three findings. First, both RP and ALT expose a coordination failure invisible to traditional metrics: Q-learning agents perform worse than random policies by 10-73% on RP and 7-35% on CALT, while Reward Fairness remains misleadingly high (above 0.92 for n>=3). Second, RP achieves 12-25x computational speedup over ALT, growing with n. Third, the two families are complementary: ALT provides richer discrimination for small populations; RP scales reliably where ALT becomes intractable. Together they form a diagnostic toolkit for temporal fair division.

Despite rapid advancements, current text-to-image (T2I) models predominantly rely on a single-step generation paradigm, which struggles with complex semantics and faces diminishing returns from parameter scaling. While recent multi-step reasoning approaches show promise, they are hindered by ungrounded planning hallucinations lacking verification, monolithic post-hoc reflection, long-context optimization instabilities, and prohibitive inference latency. To overcome these bottlenecks, we propose the Closed-Loop Visual Reasoning (CLVR) framework, a comprehensive system that deeply couples visual-language logical planning with pixel-level diffusion generation. CLVR introduces an automated data engine with step-level visual verification to synthesize reliable reasoning trajectories, and proposes Proxy Prompt Reinforcement Learning (PPRL) to resolve long-context optimization instabilities by distilling interleaved multimodal histories into explicit reward signals for accurate causal attribution. Furthermore, to mitigate the severe latency bottleneck caused by iterative denoising, we propose $Δ$-Space Weight Merge (DSWM), a theoretically grounded method that fuses alignment weights with off-the-shelf distillation priors, reducing the per-step inference cost to just 4 NFEs without requiring expensive re-distillation. Extensive experiments demonstrate that CLVR outperforms existing open-source baselines across multiple benchmarks and approaches the performance of proprietary commercial models, unlocking general test-time scaling capabilities for complex visual generation.

Generating adversarial examples at scale is a core primitive for robustness evaluation, adversarial training, and red-teaming, yet even "fast" attacks such as FGSM remain throughput-limited by the cost of a backward pass. We introduce a family of attacks that eliminates the backward pass by predicting the input gradient from forward-pass hidden states via a lightweight linear regression. The approach is motivated by a kernel view of neural networks and is exact in the Neural Tangent Kernel regime, while remaining effective for practical finite-width models. Empirically, our methods recover much of FGSM's attack performance while using only a small fraction of the time, corresponding to a $532\%$ increase in throughput. These results suggest gradient prediction as a simple and general route to significantly faster adversarial generation under realistic wall-clock constraints.

Spike activity has been the dominant neural signal for behavior decoding due to its high spatial and temporal resolution. However, as brain-computer interfaces (BCIs) move toward high channel counts and wireless operation, the high sampling frequency of spike signals becomes a bottleneck due to high power and bandwidth requirements. Local field potentials (LFPs) represent a different spatial-temporal scale of brain activity compared to spikes, offering key advantages including improved long-term stability, reduced energy consumption, and lower bandwidth requirement. Despite these benefits, LFP-based decoding models typically show reduced accuracy and often rely on non-causal architectures that are unsuitable for real-time deployment. To address these challenges, we propose REALM: a retrospective distillation framework that enables causal LFP decoding. Inspired by offline-to-online distillation strategies in speech recognition, REALM transfers representational knowledge from a pretrained multi-session bidirectional LFP model to a causal version for real-time deployment. We first pretrain a bidirectional Mamba-2 teacher model using a masked autoencoding objective. We then distill this teacher model into a compact student model via a combined objective of representation alignment and task supervision. REALM consistently outperforms both causal and non-causal LFP-based SOTA methods for behavior decoding. Notably, our REALM improves decoding performance while achieving a $2\times$ reduction in parameter count and a $10\times$ reduction in training time. These results demonstrate that retrospective distillation effectively bridges the gap between offline and real-time neural decoding. REALM shows that LFP-only models can achieve competitive decoding performance without reliance on spike signals, offering a practical and scalable alternative for next-generation wireless implantable BCIs.

As modern microservice systems grow increasingly complex due to dynamic interactions and evolving runtime environments, they experience failures with rising frequency. Ensuring system reliability therefore critically depends on accurate root cause localization (RCL). While numerous traditional machine learning and deep learning approaches have been explored for this task, they often suffer from limited interpretability and poor transferability across deployments. More recently, large language model (LLM)-based methods have been proposed to address these issues. However, existing LLM-based approaches still face two fundamental limitations: context explosion, which dilutes critical evidence and degrades localization accuracy, and serial reasoning structures, which hinder deep causal exploration and impair inference efficiency. In this paper, we conduct a comprehensive study of both how human SREs perform root cause localization in practice and why existing LLM-based methods fall short. Motivated by these findings, we introduce RCLAgent, an in-depth root cause localization framework for microservice systems that realizes multi-agent recursion-of-thought with parallel reasoning. RCLAgent decomposes the diagnostic process along the trace graph by assigning each span to a Dedicated Agent and organizing agents recursively and in parallel according to the graph topology, with the final diagnosis obtained by synthesizing the Root-Level Diagnosis Report and the Global Evidence Graph. Extensive experiments on multiple public benchmarks demonstrate that RCLAgent consistently outperforms state-of-the-art methods in both localization accuracy and inference efficiency.

AI agents execute complex multi-step processes, but current evaluation falls short: outcome metrics report success or failure without explaining why, and process-level approaches struggle to connect failure types to their precise locations within long, structured traces. We present a holistic agent evaluation framework that pairs top-down agent-level diagnosis with bottom-up span-level evaluation, decomposing analysis into independent per-span assessments. This decomposition scales to traces of arbitrary length and produces span-level rationales for each verdict. On the TRAIL benchmark, our framework achieves state-of-the-art results across all metrics on both GAIA and SWE-Bench, with relative gains over the strongest prior baselines of up to 38% on category F1, up to 3.5x on localization accuracy, and up to 12.5x on joint localization-categorization accuracy. Per-category analysis shows our framework leading in more error categories than any other evaluator. Notably, the same frontier model achieves several times higher localization accuracy when used inside our framework than as a monolithic judge over the full trace, showing that evaluation methodology, not model capability, is the bottleneck.

Training deep neural networks at scale can benefit from domain decomposition, where the network is split into subdomains trained in parallel and coupled by a global trust-region mechanism. Building on the Additively Preconditioned Trust-Region Strategy (APTS), we propose a non-monotone variant with a nonlinear additive Schwarz preconditioner that combines parallel subdomain corrections with global coarse-space directions. A windowed acceptance criterion allows controlled objective increases, avoiding needless rejection of effective coarse steps. The resulting non-monotone APTS (NAPTS) preserves accuracy while reducing CPU time by 30\% and cutting rejected steps to one third of those in APTS.

As coding agents gain access to shells, repositories, and user files, least-privilege authorization becomes a prerequisite for safe deployment: an agent should receive enough authority to complete the task, without unnecessary authority that exposes sensitive surfaces.To study whether current models can infer this boundary themselves, we first introduce permission-boundary inference, where a model maps a task instruction and terminal environment to a file-level read/write/execute policy, and AuthBench, a benchmark of 120 realistic terminal tasks with human-reviewed permission labels and executable validators for utility and attack outcomes.AuthBench shows that authorization is not a simple conservative-versus-permissive calibration problem: frontier models often omit permissions required by the execution chain while also granting unused or sensitive accesses.Increasing inference-time reasoning does not resolve this mismatch. Instead, each model moves toward a model-specific authorization attractor: more reasoning makes it more consistent in its own failure mode, whether broad-but-exposed or tight-but-brittle.This suggests that direct policy generation is the bottleneck, because a single generation must both discover all necessary accesses and reject all unnecessary ones.We therefore propose Sufficiency-Tightness Decomposition, which first generates a coverage-oriented policy by forward-simulating the task and then audits each granted entry for grounding and sensitivity.Across tested models, this decomposition improves sensitive-task success by up to 15.8% on tightness-biased models while reducing attack success across all evaluated models.

Harmonized System (HS) tariff classification is a high-stakes, expert-level task in which a free-form product description must be mapped to a specific six- or eight-digit code under the General Interpretive Rules (GIR), section notes, chapter notes, and Explanatory Notes. The difficulty lies not in knowledge volume but in *multi-dimensional rule reasoning*: a correct classification must satisfy competing priority rules along several axes simultaneously, including material, form, function, essential character, the part-versus-whole boundary, and specific listing versus residual headings. End-to-end prompting of large language models fails characteristically by resolving one axis while ignoring the priority constraints on the others. We present a *deterministic agentic workflow* in contrast to self-planning agents: the control flow is fixed, language model calls are confined to narrow stages, and reflection and verification are retained as local mechanisms. This design yields interpretability by construction--each decision is decomposed into stage-wise structured outputs with verbatim citation of the chapter or section notes that bear on it. The architecture combines offline knowledge-engineering of the Chinese HS tariff with an online six-stage pipeline. Evaluated on HSCodeComp at the six-digit level, the workflow reaches 75.0% top-1 and 91.5% top-3 at four digits, and 64.2% top-1 and 78.3% top-3 at six digits with Qwen3.6-plus; an open-weight Qwen3.6-27B-FP8 backbone in non-thinking mode achieves 84.2% four-digit and 77.4% six-digit top-1 agreement with the frontier model. A two-stage manual audit of 226 six-digit disagreements suggests that a non-trivial fraction of HSCodeComp ground-truth labels may deviate from HS general rules; full adjudication records are released in the appendix as preliminary findings for community review.

Forecasting within signal processing pipelines is crucial for mitigating delays, particularly in predicting the dynamic movements of objects such as NBA players. This task poses significant challenges due to the inherently interactive and unpredictable nature of sports, where abrupt changes in velocity and direction are prevalent. Traditional approaches, including (S)ARIMA(X), Kalman filters (KF), and Particle filters (PF), often struggle to model the non-linear dynamics present in such scenarios. Machine learning (ML) methods, such as long short-term memory (LSTM) networks, graph neural networks (GNNs), and Transformers, offer greater flexibility and accuracy but frequently fail to explicitly capture the interplay between temporal dependencies and contextual interactions, which are critical in chaotic sports environments. In this paper, we evaluate these models and assess their strengths and weaknesses. Experimental results reveal key performance trade-offs across input history length, generalizability, and the ability to incorporate contextual information. ML-based methods demonstrated substantial improvements over linear models across forecast horizons of up to 2s. Among the tested architectures, our hybrid LSTM augmented with contextual information achieved the lowest final displacement error (FDE) of 1.51m, outperforming temporal convolutional neural network (TCNN), graph attention network (GAT), and Transformers, while also requiring less data and training time compared to GAT and Transformers. Our findings indicate that no single architecture excels across all metrics, emphasizing the need for task-specific considerations in trajectory prediction for fast-paced, dynamic environments such as NBA gameplay.

Human Mesh Recovery (HMR) is fundamentally ambiguous: under occlusion or weak depth cues, multiple 3D bodies can explain the same image evidence. This ambiguity is not uniform across the body, as torso pose and root structure are often relatively well constrained, whereas distal articulations such as the arms and legs are more uncertain. Building on this observation, we propose FactorizedHMR, a two-stage framework that treats these two regimes differently. A deterministic regression module first recovers a stable torso-root anchor, and a probabilistic flow-matching module then completes the remaining non-torso articulation. To make this completion reliable, we combine a composite target representation with geometry-aware supervision and feature-aware classifier-free guidance, preserving the torso-root anchor while improving single-reference recovery of ambiguity-prone articulation. We also introduce a synthetic data pipeline that provides the paired image-camera-motion supervision under diverse viewpoints. Across camera-space and world-space benchmarks, FactorizedHMR remains competitive with strong baselines, with the clearest gains in occlusion-heavy recovery and drift-sensitive world-space metrics.

Operational plan generation and verification are critical for modern complex and rapidly changing battlefield environments, yet traditional generation and verification methods still respectively face the challenges of generation infeasibility and verification insufficiency. To alleviate these limitations, we propose an Integrated Multi-Agent Framework for Generative Operational Planning and High-Fidelity Plan Verification (IFPV). IFPV consists of two tightly coupled modules: Multi-Perspective Hierarchical Agents (MPHA) for generative operational planning and an Adversarial Cognitive Simulation Engine (ACSE) for high-fidelity adversarial plan verification. MPHA decomposes commander intent into executable multi-platform tactical action sequences through the collaboration of Pathfinder, Analyst, and Planner agents. ACSE introduces an opponent equipped with a customized world model, which predicts the future evolution of mission-critical platforms and conducts dynamic counteractions against candidate plans. Simulation experiments in the Asymmetric Combat Tactic Simulator (ACTS) show that IFPV improves mission success by 19.4% and reduces operational cost by 41.7% compared with a single-step large language model (LLM) planning baseline. Compared with a traditional rule-based validator, ACSE increases the average suppression rate by 31.8%, indicating that the proposed verification environment is stricter and more discriminative in revealing the latent vulnerabilities of candidate plans. The code for IFPV can be found at https://github.com/zhigao3ks/IFPV.

We introduce XFP, a dynamic weight quantizer for LLM inference that inverts the conventional workflow: the operator specifies reconstruction quality floors on per-channel cosine similarity (one strict floor for attention and shared experts, one lazy floor for routed-expert MoE); XFP determines codebook size, outlier budget, and packing per layer automatically -- no Hessian, no calibration data, no manual bit-width selection. Each weight matrix is decomposed into a sparse fp16 outlier residual and a dense sub-byte index tensor into a per-group learned codebook. Two storage modes share one auto-select frontend and one fused decode kernel: V2 (per-channel Lloyd) and V2a (shared library of L=32 codebooks per layer). On Qwen3.5-122B-A10B under V2, XFP reaches 138 tok/s single-stream decode on workstation hardware (RTX PRO 6000 Blackwell, TP=2) at 94.49% GSM8K strict-match (3 seeds, n=3957), and is 49% faster than Marlin INT4 at TP=1. For models that do not fit in the target memory envelope, we present the H-Process: a quality-driven iteration over the two cosine thresholds that finds the operating point at which the model just fits while still producing sensible output. Three constraints define its search space: the operator-set thresholds, an OOM boundary at quantize-on-load, and a garbage boundary in generation (cosine similarity steers; benches verify). On Qwen3.5-397B-A17B (512 routed experts/layer), the H-Process fits the full expert population into 2x96 GB at ~3.4 effective bits and delivers 100.9 tok/s long-output decode at 66.72% GSM8K strict-match on the full 1319-problem set (single seed at submission; multi-seed evaluation in progress), exceeding INT4 with routed-expert pruning on memory, throughput, and accuracy simultaneously.

Low-rank adaptation (LoRA) has become the dominant paradigm for parameter-efficient fine-tuning (PEFT) of large language models (LLMs). However, its bilinear structure introduces a critical limitation: the mapping from trainable parameters to weight updates is not distance-preserving, distorting the optimization landscape. Methods that project a low-dimensional vector into LoRA's parameter space, such as Uni-LoRA, improve parameter efficiency, but the subsequent bilinear LoRA map breaks end-to-end isometry, leaving the core distance-preservation problem unresolved. We propose GPart (Global Partition fine-tuning), a highly parameter-efficient fine-tuning method which removes the low-rank bottleneck entirely. Our method uses a single isometric partition matrix to map a $d$-dimensional trainable vector directly into the full weight space of the model. The result is an extremely minimal fine-tuning pipeline: one random projection, end-to-end isometric, with a single clean hyperparameter ($d$) and storage cost of $d+1$ values (the trainable vector plus a random seed). GPart builds on the theoretical premise that effective fine-tuning can emerge from random low-dimensional subspaces of the full weight space, without imposing low-rank matrix structure. We empirically demonstrate the superior or comparable performance of GPart to existing PEFT methods on natural language understanding, computer vision tasks, and mathematical reasoning. Overall, GPart achieves state-of-the-art efficiency and performance by removing structural constraints, offering a straightforward and elegant path to PEFT.

We study inventory control with decision-dependent censoring, focusing on the censored or repeated newsvendor (R-NV), where each order quantity determines whether demand is fully observed or censored by sales. Existing approaches based on parametric Thompson sampling (TS) can be brittle under prior mismatch, while offline imputation methods need not transfer to online learning. Motivated by the predictive view of decision making, we combine these ideas by taking oracle actions on learned completions of latent demand. We propose in-context generative posterior sampling (ICGPS), which uses modern generative models that are meta-trained offline and deployed online by in-context autoregressive generation. Theoretically, we show that the Bayesian regret of ICGPS with a learned completion kernel is bounded by the Bayesian regret of a TS benchmark with the ideal completion kernel plus a deployment penalty scaling as $\sqrt{T}$ times the square root of the completion mismatch. This yields a plug-in template for operational problems with known TS regret bounds. For R-NV, we derive sublinear Bayesian regret by reducing censored feedback to bandit convex optimization feedback. We also show that, under reasonable coverage and stability assumptions, the online completion mismatch is controlled by the offline censored predictive mismatch, so offline predictive quality transfers to online performance. Practically, we instantiate ICGPS with ChronosFlow, which combines a frozen time-series transformer backbone with a trainable conditional normalizing-flow head for fast censoring-consistent sampling. In benchmark experiments, ChronosFlow-ICGPS matches correctly specified TS, outperforms myopic and UCB-style baselines, and is robust to prior mismatch and distribution shift. ChronosFlow-ICGPS also performs well for the real-world SuperStore dataset, especially under heavy censoring.

Traditional methods for classifying global navigation satellite system (GNSS) jamming signals typically involve post-processing raw or spectral data streams, requiring complex and costly data transmission to cloud-based interference classification systems. In contrast, our proposed approach efficiently compresses GNSS data streams directly at the hardware receiver while simultaneously classifying jamming and spoofing attacks in real time. Given the growing prevalence of GNSS jamming, there is a critical need for real-time solutions suitable for power-constrained environments. This paper introduces a novel method for compressing and classifying GNSS jamming threats using generative artificial intelligence (GenAI), specifically variational autoencoders (VAEs), deployed on Google Edge tensor processing units (TPUs). The study evaluates various autoencoder (AE) architectures to compress and reconstruct GNSS signals, focusing on preserving interference characteristics while minimizing data size near the receiver hardware. The pipeline adapts large-scale AE models for Google Edge TPUs through 8-bit quantization to ensure energy-efficient deployment. Tests on raw in-phase and quadrature-phase (IQ) data, Fast Fourier Transform (FFT) data, and handcrafted features show the system achieves significant compression (>42x) and accurate classification of approximately 72 interference types on reconstructed signals (F2-score 0.915), closely matching the original signals (F2-score 0.923). The hardware-centric GenAI approach also substantially reduces jammer signal transmission costs, offering a practical solution for interference mitigation. Ablation studies on conditional and factorized VAEs (i.e., FactorVAE) explore latent feature disentanglement for data generation, enhancing model interpretability and fostering trust in machine learning (ML) solutions for sensitive interference applications.

Current language model systems remain fundamentally stateless across sessions, limiting their ability to personalize interactions over time. While retrieval-augmented generation and fine-tuning improve knowledge access and domain capability, they do not enable persistent understanding of individual users. We propose an emotion-attended stateful memory architecture that dynamically constructs user-specific conversational context using long-term history, emotional signals, and inferred intent at inference time. To evaluate its impact, we conducted a controlled A/B study across thirty non-scripted conversations spanning six emotionally distinct categories using the same underlying language model in both conditions. The memory-enriched condition consistently outperformed the stateless baseline across all evaluated scenarios. The largest gains were observed in memory grounding (95% improvement), plan clarity (57%), and emotional validation (34%). Results remained consistent even in emotionally adversarial conversations involving grief, distress, and uncertainty. These findings suggest that stateful emotional memory may represent a foundational infrastructure layer for hyper-personalized AI systems, though broader validation across larger and more diverse evaluations remains necessary

One of the bottlenecks on the way towards recursively self-improving systems is the challenge of interestingness: the ability to prospectively identify which tasks or data hold the potential for future progress. We formalize interestingness as an inductive heuristic for future compression progress and investigate its predictability using tools from Kolmogorov Complexity and Algorithmic Statistics. By analyzing complexity-runtime profiles under Length, Algorithmic, and Speed priors, we demonstrate that the inductive property of interestingness -- the capacity for past progress to signal future discovery -- is theoretically viable and empirically supported. We prove that expected future progress depends exponentially on the recency of the last observed breakthrough. Furthermore, we show that the Algorithmic Prior is significantly more optimistic than the Length Prior, yielding a quadratic increase in expected discovery for the same observed profile. These findings are experimentally confirmed across three diverse universal computational paradigms.

Existing clustering methods for functional data often prioritize partitioning accuracy over interpretability, making it challenging to extract meaningful insights when the data-generating process follows a specific underlying structure and an ordinal relationship among clusters is suspected. This work introduces K-Models, a novel framework that integrates ordinal constraints and estimates key underlying elements of the random process generating the observed functional profiles, improving both interpretability and structure identification. The proposed method is evaluated through simulations and real-world applications. In particular, it is tested on Region of Interest (ROI) curves, which represent reaction profiles from a reflectometric sensor monitoring biomolecular interactions, such as antigen-antibody binding. These curves represent changes in reflected light intensity over time at multiple measurement spots with immobilized antigens during analyte exposure, capturing the binding dynamics of the system. The goal is to identify intrinsic signal patterns solely from the observed dynamics, making this dataset an ideal benchmark for assessing the added interpretability of the proposed approach. By incorporating structural assumptions into the clustering process, K-Models enhances interpretability while maintaining performance comparable to state-of-the-art techniques, providing a valuable tool for analyzing functional data with an underlying ordinal structure.

Topological clustering, and its main algorithm ToMATo, is a clustering method from Topological Data Analysis (TDA) which has been applied successfully in several applications during the last few years. This is due to its high versatility, as clusters are detected from the persistent components in the sublevel sets of any user-defined function (gene expression, pixel values, etc), and efficiency, as topological clustering enjoys robustness guarantees. However, ToMATo is also limited in several ways. First, a graph on the data points needs to be provided as a hyper-parameter of the method (whose fine-tuning is left to the user). Second, ToMATo is known to be very sensitive to outlier values in the function range. Finally, and most importantly, ToMATo can only handle one function at a time, whereas it is critical to use several functions in various applications. In this article, we introduce ToMAToMP: the first topological clustering method able to handle several functions at the same time with theoretical guarantees. More specifically, we leverage a recent tool from multi-parameter persistent homology, called MMA decomposition, to design our clustering algorithm, and prove that it enjoys robustness properties. As corollaries, we show that it can be used to make ToMATo independent of graph tuning, and robust to outliers. Finally, we provide a set of numerical experiments showcasing the efficiency and quality of the clusterings produced by ToMAToMP, by showing strong improvement over non-topological and topological baselines for various datasets.

We describe the first experiment of conversion of Lexicon-Grammar tables for French verbs into the Lexical Markup Framework (LMF) format. The Lexicon-Grammar of the French language is currently one of the major sources of lexical and syntactic information for French. Its conversion into an interoperable representation format according to the LMF standard makes it usable in different contexts, thus contributing to the standardization and interoperability of natural language processing dictionaries. We briefly introduce the Lexicon-Grammar and the derived dictionaries; we analyse the main difficulties faced during the conversion; and we describe the resulting resource.

Graph prompt tuning has shown great potential in graph learning by introducing trainable prompts to enhance the model performance in conventional single-domain scenarios. Recent research has extended graph prompts to improve Graph Foundation Models (GFMs) by few-shot tuning auxiliary prompts. Despite their progress, most existing methods embed source-domain information into prompts, which serve either as input to GFMs or encoded during model pre-training. Such prompt entanglement with specific source domains and GFM pre-training strategy restricts their generalisability to other domains and different GFMs. Furthermore, existing GFM prompts merely rely on few-shot tuning for adaptation, neglecting the rich information in unlabelled target domain test data. Motivated by these insights, this paper aims to empower GFMs with pre-training-agnostic test-time graph prompt tuning, named GFMate. GFMate introduces centroid and layer prompts applied after pre-training on target domains, avoiding entanglement with specific source domains and model pre-training. In addition, a test-time complementary learning objective is devised to exploit both labelled and unlabelled target domain data for effective test-time prompt tuning. Extensive experiments on 12 benchmark datasets demonstrate the superior performance and efficiency of GFMate, achieving improvements of up to 30.63%. Code is available at https://github.com/YanJiangJerry/GFMate.

Large language models often suffer from fact loss, timeline confusion, persona drift, and reduced stability during long-range interaction, especially under high-noise knowledge bases, context clearing, and cross-model transfer. To address these issues, we introduce ARPM, an external temporal memory governance framework for long-term dialogue. ARPM separates static knowledge memory from dynamic dialogue experience memory and combines vector retrieval, BM25, RRF fusion, dual-temporal reranking, chronological evidence reading, and a controlled analysis protocol for evidence verification and answer binding. Unlike approaches that encode persona consistency into model weights or rely only on long context, ARPM treats continuity as a traceable, auditable, and transferable governance problem. Using engineering logs, we conduct three experiments. First, in a 50-round question-answering setting, we compare signal-to-noise ratios of 1:5 and 1:200+, and distinguish CSV auto-judgment from manual review. Under 1:5, CSV recall accuracy is 54.0%, while manual review raises it to 100.0%. Under 1:200+, the values are 44.0% and 80.0%. These results show that automatic rules can underestimate recall after supporting evidence enters the prompt. Second, ablation results show that dialogue history retrieval is necessary for recent continuity: disabling it reduces strict accuracy from 100% to 66.7%, and disabling BM25 reduces it to 80.0%, indicating that pure semantic retrieval is insufficient for correction and tracing. Third, under a 5.1-million-character noise substrate, periodic context clearing, and multi-model handoff, ARPM maintains semantic continuity, boundary continuity, and persona consistency, while exposing limits caused by weak protocol compliance. These findings show that long-term persona consistency can be decomposed into governable components and evaluated in a white-box manner.

Recent advances in artificial intelligence (AI) agents are pushing AI beyond tools toward autonomous scientific discovery. We discuss two complementary agentic systems for cosmology: \texttt{CMBEvolve}, which targets tasks with explicit quantitative objectives through LLM-guided code evolution and tree search, and \texttt{CosmoEvolve}, which targets open-ended scientific workflows through a virtual multi-agent research laboratory. As preliminary demonstrations, we apply \texttt{CMBEvolve} to out-of-distribution detection in weak-lensing maps, where it iteratively improves the benchmark score through code evolution, and \texttt{CosmoEvolve} to autonomous ACT DR6 data analysis, where it identifies non-trivial pair- and scale-dependent behaviour and produces analysis-grade diagnostics. These examples show how cosmology can provide both controlled benchmark tasks and realistic open-ended research problems for the development of AI scientist systems.

Research idea generation is the innovation-driving step of automated scientific research. Recently, large language models (LLMs) have shown potential for automating idea generation at scale. However, existing methods mainly condition LLMs on eliciting idea generation through static retrieval of relevant literature or complex prompt engineering, without discarding the structural relations among references. We propose Graphs of Research (GoR), a supervised fine-tuning method that extracts a 2-hop reference neighborhood for each seed paper, derives the relations among those references from citation position, frequency, predecessor links, and publication time, and organizes them into a paper-evolution directed acyclic graph (DAG). We construct an automated extraction pipeline that draws data from five major ML/NLP venues, comprising 498/50/50 train/validation/test seed papers and approximately 7,600 cited references. Qwen2.5-7B-Instruct-1M is fine-tuned on a structured-text prompt that includes the citation graph, edge signals, reference information, and task definition to predict the idea for the seed paper. Across head-to-head LLM-judge tournaments against gpt-4o-driven baselines, GoR-SFT achieves SOTA, demonstrating the effectiveness of citation-evolution graphs as supervision signal for LLM-based idea generation. We hope that this reduces the barrier for citation evolution graphs as a supervision, accelerating automated scientific innovation.

Composed Image Retrieval (CIR) is a multimodal retrieval task where a query consists of a reference image and a textual modification, and the goal is to retrieve a target image satisfying both. In principle, strong performance on CIR benchmarks is assumed to require multimodal composition, i.e., combining complementary information from reference image and textual modification. In this work, we show that this assumption does not always hold. Across four widely used CIR benchmarks and eleven Generalist Multimodal Embedding models, a large fraction of queries can be solved using a single modality (from 32.2% to 83.6%), revealing pervasive unimodal shortcuts. Thus, high CIR performance can arise from unimodal signals rather than true multimodal composition. To better understand this issue, we perform a two-stage audit. First, we identify shortcut-solvable queries through cross-model analysis. Second, we conduct human validation on 4,741 shortcut-free queries, of which only 1,689 are well-formed, with common issues including ambiguous edits and mismatched targets. Re-evaluating models on this validated subset reveals qualitatively different behaviour: queries can no longer be solved with a single modality, and successful retrieval requires combining both inputs. While accuracy decreases, reliance on multimodal information increases. Overall, current CIR benchmarks conflate shortcut-solvable, noisy, and genuinely compositional queries, leading to an overestimation of model capability in multimodal composition.

As LLM-based agents increasingly browse the web on users' behalf, a natural question arises: can websites passively identify which underlying model powers an agent? Doing so would represent a significant security risk, enabling targeted attacks tailored to known model vulnerabilities. Across 14 frontier LLMs and four web environments spanning information retrieval and shopping tasks, we show that an agent's actions and interaction timings, captured via a passive JavaScript tracker, are sufficient to identify the underlying model with up to 96\% F1. We formalise this attack surface by demonstrating that classifiers trained on agent actions generalise across model sizes and families. We further show that strong classifiers can be trained from few interaction traces and that agent identity can be inferred early within an episode. Injecting randomised timing delays between actions substantially degrades classifier performance, but does not provide robust protection: a classifier retrained on delayed traces largely recovers performance. We release our harness and a labelled corpus of agent traces \href{https://github.com/KabakaWilliam/known_actions}{here}.

Neural networks suffer from catastrophic forgetting in class-incremental learning (CIL) settings. Rehearsal$\unicode{x2013}$replaying a subset of past samples$\unicode{x2013}$is a well-established mitigation strategy. However, recent results suggest that, despite balanced rehearsal allocation, some classes are forgotten substantially more than others. Despite its relevance, this imbalanced forgetting phenomenon remains underexplored. This work shows that imbalanced forgetting arises systematically and severely in rehearsal-based CIL and investigates it extensively. Specifically, we construct, from a principled analysis, three last-layer coefficients that capture different gradient-level sources of interference affecting each past class during an incremental step. We then demonstrate that, together, they reliably predict how past classes will rank in terms of forgetting at the end of that step. While predictive performance alone does not establish causality, these results support the interpretation of the coefficients as a plausible mechanistic account linking last-layer gradient-level interactions during training to class-level forgetting outcomes. Notably, one coefficient$\unicode{x2013}$capturing self-induced interference$\unicode{x2013}$emerges as the strongest predictor, with controlled experiments providing evidence consistent with this coefficient being influenced by the new-class interference coefficient. Overall, our findings provide valuable insights and suggest promising directions for mitigating imbalanced forgetting by reducing class-wise disparities in the identified sources of interference.

In the world of AI and advanced technologies investigation aspects identification of a crime or criminal plays a major problem. In this research we focus on a Conventional ways of implicating criminal investigations usually rely on limited data analysis. Finding an optimal and efficient method that will effectively identify criminals from complex datasets and minimise false positives and false negatives is the considered as a challenge. The main novelty approach of this work is based on the deep learning algorithm Deep Deterministic Policy Gradient (DDPG) is presented in this paper. We train the DDPG model with a dataset of crime scene material, witness statements and suspect profiles. The algorithm uses features to maximise the likelihood of identifying the offender while minimising the noise impact and irrelevant data. We show the efficacy of the proposed method, where DDPG identified criminals with an amazing accuracy of 95% than other several existing methods.

Data selection accelerates training by identifying representative training data while preserving model performance. However, existing methods mainly focus on designing sample-importance criteria, i.e., deciding what to select, while typically fixing the selected data volume as the target ratio throughout training. Thus, they are often dynamic in sample identity but static in data volume. In this work, we revisit data selection from an optimization perspective and show that selected-data training induces an implicit regularization effect modulated by the instantaneous selection ratio. This reveals a key trade-off: lower ratios amplify selection-induced regularization, whereas higher ratios preserve data coverage and optimization fidelity. Motivated by this insight, we propose PODS, a Plug-and-play Oscillatory Data-volume Scheduling framework. Rather than introducing another sample-scoring metric, PODS serves as a lightweight module that dynamically schedules how much data to select over training. Under the target selection ratio, PODS alternates between low-ratio regularization phases and high-ratio recovery phases to exploit selection-induced regularization without sacrificing optimization stability. With its lightweight, ratio-level, and task-agnostic design, PODS is compatible with existing static and dynamic selection methods and broadly applicable across training paradigms. Experiments across various datasets, architectures, and tasks show that PODS consistently improves the efficiency-generalization trade-off, e.g., reducing ImageNet-1k training cost by 50% with improved accuracy and accelerating LLM instruction tuning by over 2x without performance degradation.

Understanding human motion beyond surface kinematics is crucial for motion analysis, rehabilitation, and injury risk assessment. However, progress in this domain is limited by the lack of large-scale datasets with biomechanical annotations, and by existing approaches that cannot directly infer internal biomechanical states from visual observations. In this paper, we introduce a simulation-based framework for estimating muscle activations from existing motion capture datasets, resulting in BioHuman10M, a large-scale dataset with synchronized video, motion, and activations. Building on BioHuman10M, we propose BioHuman, an end-to-end model that takes monocular video as input and jointly predicts human motion and muscle activations, effectively bridging visual observations and internal biomechanical states. Extensive experiments demonstrate that BioHuman enables accurate reconstruction of both kinematic motion and muscle activity, and generalizes across diverse subjects and motions. We believe our approach establishes a new benchmark for video-based biomechanical understanding and opens up new possibilities for physically grounded human modeling.

MediaClaw is a multimodal agent platform built on the OpenClaw ecosystem. Its core design follows a three-layer architecture of unified abstraction, pluginized extension, and workflow orchestration. The system is intended to address practical deployment pain points in AIGC adoption, including fragmented capabilities, heterogeneous interfaces, disconnected production processes, and limited reuse of high-quality production workflows. \system{} abstracts full-category AIGC capabilities into a unified invocation model, uses plugins to support hot-pluggable capability expansion, and uses task-oriented Skills to turn complex production processes into reusable workflow assets. This report focuses on the architectural design philosophy of MediaClaw, the design logic of its core capability model, and the key engineering trade-offs in implementation. It aims to provide reusable practical reference for building multimodal capability platforms.

De novo crystal generation, a central task in materials discovery, aims to generate crystals that are simultaneously valid, stable, unique, and novel. Existing methods mainly rely on black-box stochastic sampling, providing limited control over how generated structures move beyond the observed distribution. In this paper, we introduce a concept-based compositional framework for crystal generation. We train a vector-quantized variational autoencoder to automatically discover a shared set of reusable crystal concepts, which serve as building blocks for guided generation. These learned concepts naturally exhibit interpretability from both local atomic environments and global symmetry patterns, and generalize to crystals from different distributions. By recombining such concepts, our framework enables controllable exploration of novel crystals beyond the training distribution, rather than relying solely on unconstrained random sampling. To further improve composition efficiency, we introduce a composition generator and iteratively refine it using high-quality samples generated by the model itself. The resulting concept compositions are then used to condition downstream crystal generation. Numerical experiments on MP-20 and Alex-MP-20 show that compositing concepts separately increase base model up to 53.2% and 51.7% on V.S.U.N metric, with particular gains in novelty.

Normally, a system that translates speech into text consists of separate modules for speech recognition and text-to-text translation. Combining those tasks into a SpeechLLM promises to exploit paralinguistic information in the speech and to reduce cascaded errors. But existing SpeechLLM systems are slow since they do not work in a real streaming fashion: they wait for a complete utterance of audio before outputting a translation, or output tokens at fixed intervals, which is not suitable for real applications. This work proposes an LLM-based architecture for real streaming speech-to-text translation. The LLM learns not just to emit output tokens, but also to decide whether it has seen enough audio to do so. The system is trained using automatic alignments of the input speech and the output text. In experiments on different language pairs, the system achieves a translation quality close to the non-streaming baseline, but with a latency of only 1-2 seconds.

Persian music, with its unique tonalities, modal systems (Dastgah), and rhythmic structures, presents significant challenges for music generation models trained primarily on Western music. We address this gap by curating the first large-scale dataset of Persian songs, comprising over 900 hours high-quality audio samples across diverse sub-genres, including pop, traditional, and contemporary styles. This dataset captures the rich melodic and cultural diversity of Persian music and serves as the foundation for fine-tuning MusicGen, a state-of-the-art generative music model. We adapt MusicGen to this domain and evaluate its performance by utilizing subjective and objective metrics. To assess the semantic alignment between generated music and intended style tags, we report the proportion of relevant tags accurately reflected in the generated outputs. Our results demonstrate that the fine-tuned model produces compositions that more align with Persian stylistic conventions. This work introduces a new resource for generative music research and illustrates the adaptability of music generation models to underrepresented cultural and linguistic contexts.

Identifying the structural priors that enable Deep Neural Networks (DNNs) to overcome the curse of dimensionality is a fundamental challenge in machine learning theory. Existing literature suggests that effective high-dimensional learning is driven by compositional sparsity, where target functions decompose into constituents supported on low-dimensional variable subsets. To investigate this hypothesis, we combine Information Filtering Networks (IFNs), which extract sparse dependency structures via constrained information maximisation, with Homological Neural Networks (HNNs), which map the inferred topology into fixed-wiring sparse neural graphs. We formalise the design principles underlying this construction and present an interpretable pipeline in which abstraction emerges through hierarchical composition. HNNs are orders of magnitude sparser than standard DNNs and require only minimal hyperparameter tuning. On synthetic tasks with known sparse hierarchies, HNNs recover the underlying compositional structure and remain stable in regimes where dense alternatives degrade as dimensionality increases. Across a broad suite of real-world datasets, HNNs consistently match or outperform dense baselines while using far fewer parameters, exhibiting lower variance and showing reduced sensitivity to hyperparameters.

Accurately predicting individual aesthetic evaluation for images is a fundamental challenge for AI. Various deep learning (DL)-based models have been proposed for this task, training on image evaluation data to extract objective low-level features. However, aesthetic preferences are inherently subjective and individual-dependent. Accurate prediction thus requires the extraction of high-level semantic features of images and the active collection of preference information from the target individual. To address this issue, we focus on the utility of Large Language Models (LLMs) pretrained on vast amounts of textual data, and develop an integrated DL-LLM system. The system actively elicits aesthetic preferences through LLM-based semi-structured interviews and predicts aesthetic evaluation by leveraging both low-level and high-level features. In our experiments, we compare the proposed system against conventional systems, human predictors, and the target individual's own re-evaluations after a certain time interval. Our results show that the proposed system outperforms all of them, with particularly strong performance on highly-rated images. Moreover, the prediction error of the proposed system is smaller than within-person variability, while human predictors show the largest error, likely due to the influence of their own aesthetic values. These results suggest that AI may be better positioned than others or one's future self to capture individual aesthetic preferences at a given point. This opens a new question of whether AI could serve as a deeper interpreter of human aesthetic sensibility than humans themselves.

De novo crystal generation seeks to discover materials that are not merely realistic, but also stable and novel. However, most existing generative models are trained to maximize the likelihood of observed crystals, which encourages samples to stay close to known materials yet not necessarily align with the criteria that matter in discovery. Through an empirical investigation, we show that current crystal generative models are caught in a pronounced stability--novelty trade-off: moving toward the observed distribution preserves stability but limits novelty, whereas moving away from it quickly destroys stability. This suggests that the useful region for discovering crystals that are both stable and novel is extremely narrow. To escape the trade-off, we introduce Crys-JEPA, a joint embedding predictive architecture for crystals that learns an energy-aware latent space preserving formation-energy differences. In this space, stability assessment can be reformulated as an embedding-based comparison against accessible training crystals, reducing the reliance on expensive energy evaluation and task-specific external references. Building on Crys-JEPA, we further develop a screening-and-refinement pipeline that identifies promising generated crystals and reintroduces them to refine the generative model. On MP-20 and Alex-MP-20 datasets, we achieve improvements over baselines up to 81.4% and 82.6% on V.S.U.N metric, respectively.

History-dependent policies induced by recurrent neural networks (RNNs) rely on latent hidden state dynamics, making verification in partially observable reinforcement learning (RL) challenging. Existing RNN verification tools typically rely on restrictive modeling assumptions or coarse over-approximations of the hidden state space, which can lead to overly conservative or inconclusive results. We propose $\textbf{RNN}$ $\textbf{Pro}$babilistic $\textbf{Ve}$rification ($\texttt{RNN-ProVe}$), a probabilistic framework that $\textit{estimates the likelihood}$ of undesired behaviors in RNN-based policies. $\texttt{RNN-ProVe}$ uses policy-driven sampling to approximate the set of hidden states that are feasible under a trained policy, and derives statistical error bounds to produce bounded-error, high-confidence estimates of behavioral violations. Experiments on partially observable single-agent and cooperative multi-agent tasks show that $\texttt{RNN-ProVe}$ yields more quantitative, feasibility-aware probabilistic guarantees than existing tools, while scaling to recurrent and multi-agent settings.

Large Language Models (LLMs) are increasingly deployed for knowledge synthesis, yet their capacity for compositional generalization in scientific knowledge remains under-characterized. Existing benchmarks primarily focus on single-turn restricted scenarios, failing to capture the capability boundaries exposed by real-world interactive scientific workflows. To address this, we introduce XDomainBench, a diagnostic benchmark for interactive interdisciplinary scientific reasoning. We formalize the composition order and mixture structure to enable systematic stress-testing from single-discipline to inter-disciplinary, comprising 8,598 interactive sessions across 20 domains and 4 task categories, with 8 realistic trajectory patterns covering difficulty and domain-mixture dynamics, simulating real AI4S scenarios. Large-scale evaluation of LLMs reveals a systematic reasoning collapse as composition order increases, stemming from two root causes: (i) direct difficulty increases induced by domain composition, and (ii) indirect interaction-amplified failures where trajectory patterns trigger error accumulation, reasoning breaks, and domain confusion, ultimately leading to session collapse.

Accurately identifying student misconceptions is crucial for personalized education but faces three challenges: (1) data scarcity with long-tail distribution, where authentic student reasoning is difficult to synthesize; (2) fuzzy boundaries between error categories with high annotation noise; (3) deployment parado-large models overlook unconventional approaches due to pretraining bias and cannot be deployed on edge, while small models overfit to noise. Unlike traditional methods that increase diversity through large-scale data synthesis, we propose a two-stage knowledge distillation framework that mines high-value samples from existing data. The first stage performs standard distillation to transfer task capabilities. The second stage introduces a dual-layer marginal selection mechanism based on cognitive uncertainty, identifying four types of critical samples based on teacher model uncertainty and confidence differences. For different data subsets, we design difficulty-adaptive mechanism to balance hard/soft label contributions, enabling student models to inherit inter-class relationships from teacher soft labels while distinguishing ambiguous error types. Experiments show that with augmented training on only 10.30% of filtered samples, we achieve MAP@3 of 0.9585 (+17.8%) on the MAP-Charting dataset, and using only a 4B parameter model, we attain 84.38% accuracy on cross-topic tests of middle school algebra misconception benchmarks, significantly outperforming sota LLM (67.73%) and standard fine-tuned 72B models (81.25%). Our code is available at https://github.com/RoschildRui/acl2026_map.

Large Language Models (LLMs) and Vision Language Models (VLMs) have demonstrated impressive capabilities but remain vulnerable to jailbreaking attacks, where adversaries exploit textual or visual triggers to bypass safety guardrails. Recent defenses typically rely on safety fine-tuning or external filters to reduce the model's likelihood of producing harmful content. While effective to some extent, these methods often incur significant computational overheads and suffer from the safety utility trade-off, degrading the model's performance on benign tasks. To address these challenges, we propose EVA (Editing for Versatile Alignment against Jailbreaks), a novel framework that pioneers the application of direct model editing for safety alignment. EVA reframes safety alignment as a precise knowledge correction task. Instead of retraining massive parameters, EVA identifies and surgically edits specific neurons responsible for the model's susceptibility to harmful instructions, while leaving the vast majority of the model unchanged. By localizing the updates, EVA effectively neutralizes harmful behaviors without compromising the model's general reasoning capabilities. Extensive experiments demonstrate that EVA outperforms baselines in mitigating jailbreaks across both LLMs and VLMs, offering a precise and efficient solution for post-deployment safety alignment.

Intervention is one of the most representative and widely used methods for understanding the internal representations of large language models (LLMs). However, existing intervention methods are confined to linear interventions grounded in the Linear Representation Hypothesis, leaving features encoded along non-linear manifolds beyond their reach. In this work, we introduce a general formulation of intervention that extends naturally to non-linearly represented features, together with a learning procedure that further enables intervention on implicit features lacking a direct output signature. We validate our framework on refusal bypass steering, where it steers the model more precisely than linear baselines by intervening on a non-linear feature governing refusal.

Recent advances in multimodal large language models have driven growing interest in graphical user interface (GUI) agents, yet their generalization remains constrained by the scarcity of large-scale training data spanning diverse real-world applications. Existing datasets rely heavily on costly manual annotations and are typically confined to narrow domains. To address this challenge, we propose Video2GUI, a fully automated framework that extracts grounded GUI interaction trajectories directly from unlabeled Internet videos. Video2GUI employs a coarse-to-fine filtering strategy to identify high-quality GUI tutorial videos and convert them into structured agent trajectories. Applying this pipeline to 500 million video metadata entries, we construct WildGUI, a large-scale dataset containing 12 million interaction trajectories spanning over 1,500 applications and websites. Pre-training Qwen2.5-VL and Mimo-VL on WildGUI yields consistent improvements of 5-20% across multiple GUI grounding and action benchmarks, matching or surpassing state-of-the-art performance. We will release both the WildGUI dataset and the Video2GUI pipeline to support future research of GUI agents.

While large language models (LLMs) are trained to align with human values, their generations may still violate safety constraints. A growing line of work addresses this problem by modifying the model's sampling policy at decoding time using a safety reward. However, existing decoding-time steering methods often intervene unnecessarily, modifying generations that would have been safe under the base model. Such unnecessary interventions are undesirable, as they can distort key properties of the base model such as helpfulness, fluency, style, and coherence. We propose a new test-time steering method designed to reduce such unnecessary interventions while improving the safety of unsafe responses. Our approach filters tokens using a value-based safety criterion and provides an explicit bound on the probability of false interventions. A single threshold hyperparameter controls this bound, allowing practitioners to trade off higher rates of unnecessary intervention for better output safety. Across multiple datasets and experiments, we show that our value-filtered decoding method outperforms existing baselines, achieving better trade-offs between safety, helpfulness, and similarity to the base model.

Large language models in regulated financial workflows are governed by natural-language policies that the same model interprets, creating a principal--agent failure: outputs can appear compliant without being compliant. Existing evaluation measures task accuracy but not whether governance constrains behaviour at the decision rationale level -- where regulated decisions must be auditable. We introduce five governance metrics that quantify policy compliance at the rationale level and apply them in a synthetic banking domain to compare text-only governance against mechanical enforcement: four primitives operating outside the model's interpretive loop. Under text-only governance, 27% of deferrals carry no decision-relevant information. Mechanical enforcement reduces this rate by 73%, more than doubles deferral information content, and raises task accuracy from MCC~$0.43$ to $0.88$. The improvement is driven by architectural separation: LLM-generated rationales under mechanical enforcement show comparable CDL to text-only governance -- the gain comes from removing clear-cut decisions from the model's control. A causal ablation confirms that each primitive is individually necessary. Our central finding is a governance-task decoupling: under structural stress, text-only governance degrades on both dimensions simultaneously, whereas mechanical enforcement preserves governance quality even as task performance drops. This implies that governance and task evaluation are distinct axes: accuracy is not a sufficient proxy for governance in regulated AI systems.

Electrified chemical processes are incentivized by exposure to time-varying electricity markets to operate flexibly, but participating in demand response schemes can require satisfying terminal constraints over long horizons. Specifically, terminal constraints may be required when computing optimal schedules in order to preserve dynamic stability. Model-based optimization methods are computationally costly, and data-driven scheduling via reinforcement learning (RL) faces severe credit-assignment challenges. We integrate Goal-Space Planning (GSP) with Deep Deterministic Policy Gradient (DDPG), using learned temporally abstract models over discrete subgoals to propagate value across extended horizons. Using a simulated air separation benchmark, we demonstrate the proposed approach improves sample efficiency over standard DDPG while satisfying terminal storage constraints, mitigating myopic control behavior.

Recent work has promoted task-aware layer pruning as a way to improve model performance on particular tasks, as shown by TALE. In this paper, we investigate when such improvements occur and why. We show first that, across controlled polynomial regression tasks and large language models, such pruning yields no benefit on in-distribution (ID) data but consistently improves out-of-distribution (OOD) accuracy. We further show empirically that OOD inputs induce layerwise norm and pairwise-distance profiles that deviate from the corresponding ID profiles. This leads to a geometric explanation of task-aware pruning: each task induces a task-adapted geometry, characterized empirically by the representation profiles observed on ID inputs. OOD inputs can introduce a distorted version of the task-adapted geometry. Task-aware pruning identifies layers that create or amplify this distortion; by removing them, it shifts OOD representational norms and pairwise distances toward those observed on the adapted distribution. This realigns OOD inputs with the model's task-adapted geometry and improves performance. We provide causal evidence through controlled distribution shifts and residual-scaling interventions, and demonstrate consistent behavior across model scales.

Target speech extraction remains difficult for compact devices because monaural neural models lack spatial evidence and classical beamformers lose resolving power when the microphone aperture is only a few centimetres. We present IsoNet, a user-selectable audio-visual target speech extraction system for a compact 4-microphone array. IsoNet combines complex multi-channel STFT features, GCC-PHAT spatial cues, face-conditioned visual embeddings, and auxiliary direction-of-arrival supervision inside a U-Net mask estimation network. Three curriculum variants were trained on 25,000 simulated VoxCeleb mixtures with progressively difficult SNR regimes. On a hard test set spanning -1 to 10 dB SNR, IsoNet-CL1 achieves 9.31 dB SI-SDR, a 4.85 dB improvement over the mixture, with PESQ 2.13 and STOI 0.84. Oracle delay-and-sum and MVDR beamformers degrade the same mixtures by 4.82 dB and 6.08 dB SI-SDRi, respectively, showing that the proposed learned multimodal conditioning solves a regime where conventional spatial filtering is ineffective. Ablation studies show consistent gains from visual conditioning, GCC-PHAT features, and extended delay-bin encoding. The results establish a compact-array, face-selectable speech extraction baseline under controlled simulation and identify the remaining barriers to real deployment, especially phase reconstruction, multi-interferer mixtures, and simulation-to-real transfer.

Sepsis management in the ICU requires sequential treatment decisions under rapidly evolving patient physiology. Although large language models (LLMs) encode broad clinical knowledge and can reason over guidelines, they are not inherently grounded in action-conditioned patient dynamics. We introduce SepsisAgent, a world model-augmented LLM agent for sepsis treatment recommendation. SepsisAgent uses a learned Clinical World Model to simulate patient responses under candidate fluid--vasopressor interventions, and follows a propose--simulate--refine workflow before committing to a prescription. We first show that world-model access alone yields inconsistent LLM decision performance, motivating agent-specific training. We then train SepsisAgent through a three-stage curriculum: patient-dynamics supervised fine-tuning, propose--simulate--refine behavior cloning, and world-model-based agentic reinforcement learning. On MIMIC-IV sepsis trajectories, SepsisAgent outperforms all traditional RL and LLM-based baselines in off-policy value while achieving the best safety profile under guideline adherence and unsafe-action metrics. Further analysis shows that repeated interaction with the Clinical World Model enables the agent to learn regularities in patient evolution, which remain useful even when simulator access is removed.

Strong equivalence between knowledge bases ensures the possibility of replacing one with the other without affecting reasoning outcomes, in any given context. This makes it a crucial property in nonmonotonic formalisms. In particular, the fields of logic programming and abstract argumentation provide primary examples in which this property has been subject to vast investigations. However, while (classes of) logic programs and abstract argumentation frameworks are known to be semantically equivalent in static settings, this alignment breaks in dynamic contexts due to differing notions of update. As a result, strong equivalence does not always carry over from one formalism to the other. In this paper, we carefully investigate this discrepancy and introduce a new notion of strong equivalence for logic programs. Our approach preserves strong equivalence under translation between certain classes of logic programs and both Dung-style and claim-augmented argumentation frameworks, thus restoring compatibility across these formalisms.

Label-free single-cell imaging offers a scalable, non-invasive alternative to fluorescence-based cytometry, yet inferring molecular phenotypes directly from bright-field morphology remains challenging. We present a unified Deep Learning (DL) framework that jointly performs White Blood Cell (WBC) classification and continuous protein-expression regression from label-free Differential Phase Contrast (DPC) images. Our model employs a Hybrid architecture that fuses convolutional fine-grained texture features with transformer-based global representations through a learnable cross-branch gating module, enabling robust morpho-molecular inference from DPC images. To support downstream interpretability, we further incorporate a Large Language Model (LLM) that generates concise, biologically grounded summaries of the predicted cell states. Experiments on the Berkeley Single Cell Computational Microscopy (BSCCM) and Blood Cells Image benchmarks demonstrate strong performance, achieving a 91.3% WBC classification accuracy and a 0.72 Pearson correlation for CD16 expression regression on BSCCM. These results underscore the promise of label-free single-cell imaging for cost-effective hematological profiling, enabling simultaneous phenotype identification and quantitative biomarker estimation without fluorescent staining. The source code is available at https://github.com/saqibnaziir/Single-Cell-Phenotyping.

Sparse anchors provide a compact interface for human motion authoring: users specify a few root positions, planar trajectory samples, or body-point targets, while the system synthesizes the full-body motion that completes the under-specified intent. We present AnchorRoute, a sparse-anchor motion synthesis framework that uses anchors as a shared scaffold for both generation and refinement. Before generation, AnchorRoute converts sparse anchors into anchor-condition features and injects the resulting condition memory into a frozen Transition Masked Diffusion prior through AnchorKV and dual-context conditioning. This preserves the generation quality of the pretrained text-to-motion prior while learning sparse spatial control. After generation, the same anchors are evaluated as residuals: their timestamps define refinement intervals, and their residuals determine where correction should be concentrated. RouteSolver then refines the motion by projecting soft-token updates onto anchor-defined piecewise-affine interval bases. This couples generation-time anchor conditioning with residual-routed refinement under one anchor scaffold. AnchorRoute supports root-3D, planar-root, and body-point control within the same formulation. In benchmark evaluations, AnchorRoute outperforms prior sparse-control methods under the sparse keyjoint protocol and consistently improves anchor adherence across control families. The results show that the learned anchor-conditioned generator and RouteSolver refinement are complementary: the generator preserves text-motion quality, while RouteSolver provides a controllable path toward stronger anchor adherence.

Robot imitation data are often multimodal: similar visual-language observations may be followed by different action chunks because human demonstrators act with different short-horizon intents, task phases, or recent context. Existing frame-conditioned VLA policies infer each chunk from the current observation and instruction alone, so under partial observability they may resample different intents across adjacent replanning steps, leading to inter-chunk conflict and unstable execution. We introduce IntentVLA, a history-conditioned VLA framework that encodes recent visual observations into a compact short-horizon intent representation and uses it to condition chunk generation. We further introduce AliasBench, a 12-task ambiguity-aware benchmark on RoboTwin2 with matched training data and evaluation environments that isolate short-horizon observation aliasing. Across AliasBench, SimplerEnv, LIBERO, and RoboCasa, IntentVLA improves rollout stability and outperforms strong VLA baselines

Deep learning and multi-modal fusion have demonstrated transformative potential in medical diagnosis by integrating diverse data sources. However, accurate prognosis for ischemic stroke remains challenging due to limitations in existing multi-modal approaches. First, current methods are predominantly confined to dual-modal fusion, lacking a framework that effectively integrates the trifecta of medical images, structured clinical data, and unstructured text. Second, they often fail to establish deep bidirectional interactions between modalities; To address these critical gaps, this paper proposes a novel tri-modal fusion model for ischemic stroke prognosis. Our approach first enriches the data representation by employing a Large Language Model (LLM) to automatically generate semi-structured diagnostic text from brain MRIs. This process not only addresses the scarcity of expert annotations but also serves as a regularized semantic enhancement, improving multimodal fusion robustness. Furthermore, we design a core component termed the Vision-Conditioned Dual Alignment Fusion Module (VDAFM), which strategically uses visual features as a conditional prior to guide fine-grained interaction with the generated text. This module achieves a dynamic and profound fusion through a dual semantic alignment loss, effectively mitigating modal heterogeneity. Extensive experiments on a real-world clinical dataset demonstrate that our model achieves state-of-the-art performance.

In real-world scenes, target objects may reside in regions that are not visible. While humans can often infer the locations of occluded objects from context and commonsense knowledge, this capability remains a major challenge for vision-language models (VLMs). To address this gap, we introduce SceneFunRI, a benchmark for Reasoning the Invisible. Based on the SceneFun3D dataset, SceneFunRI formulates the task as a 2D spatial reasoning problem via a semi-automatic pipeline and comprises 855 instances. It requires models to infer the locations of invisible functional objects from task instructions and commonsense reasoning. The strongest baseline model (Gemini 3 Flash) only achieves an CAcc@75 of 15.20, an mIoU of 0.74, and a Dist of 28.65. We group our prompting analysis into three categories: Strong Instruction Prompting, Reasoning-based Prompting, and Spatial Process of Elimination (SPoE). These findings indicate that invisible-region reasoning remains an unstable capability in current VLMs, motivating future work on models that more tightly integrate task intent, commonsense priors, spatial grounding, and uncertainty-aware search.

Foundation models (FMs) promise to extract unified representations that generalize across downstream tasks. They have emerged across fields, including electroencephalography (EEG), but it is less clear how effective they are in this particular field. Published evaluations differ in datasets, in the EEG-specific preprocessing that might influence reported results, and in the reported metrics, frequently obscuring the clinical relevance in EEG. We introduce NeuroAtlas, the largest EEG benchmark to date: 42 datasets and 260k hours covering clinical EEG (epilepsy, sleep medicine, brain age estimation) and brain-computer interfaces, and include multiple datasets per task along with bespoke clinical evaluation metrics. Besides evaluating EEG-FMs with respect to supervised baselines, we present results from generic time-series FMs. We report three findings. First, EEG-specific FMs do not consistently outperform time-series FMs, which have neither EEG-focused architectures nor been pretrained on EEG. Second, standard machine learning metrics are insufficient to assess clinical utility: thus, we thoroughly evaluate more appropriate measures such as the quality of event-level decision-making, hypnogram-derived features, and the brain-age gap in the domains of epilepsy, sleep, and brain age, respectively. Third, model rankings and performance can vary substantially within domains. We conclude that pretrained models perform largely on par, with only narrow advantages for a few, and that current models do not yet deliver on the promise of an out-of-the-box unified EEG model. NeuroAtlas exposes this gap and provides the datasets and metrics for the next generation of unified EEG FMs.

Sparse Autoencoders (SAEs) that can accurately reconstruct their input (minimizing distortion) by making efficient use of few features (minimizing the rate) often fail to learn monosemantic representations (highly interpretable), limiting their usefulness for mechanistic interpretability. In this paper, we characterise this tension in learning faithful, efficient, and interpretable explanations, introducing the Rate-Distortion-Polysemanticity tradeoff in SAEs. Under toy-modeling assumptions, we theoretically and empirically show that restricting the SAE to be monosemantic necessarily comes with an increase in rate and distortion. Assuming a generative model behind the input observations, we further demonstrate that the degree of polysemanticity of optimal SAEs is determined by the training data distribution, especially by the probability of features to co-occur. Finally, we extend the analysis to real-world settings by deriving necessary conditions that a polysemanticity measure should satisfy when the data-generating process is unknown, and we benchmark existing proxy metrics on SAEs trained on Large Language Models. Taken together, our findings show that polysemanticity is a data problem that should be accounted for when addressing it at the architectural and optimization level.

Tabular data sharing under privacy constraints is increasingly important for research and collaboration. Synthetic data generators (SDGs) are a promising solution, but synthetic data remains vulnerable to attacks, such as membership inference attacks (MIAs), which aim to determine whether a specific record was part of the training data. State-of-the-art MIAs are powerful but impractical: they rely on shadow modeling, requiring hundreds of SDG training runs, and need auxiliary data several times larger than the original training set. Fast proxy metrics like distance to closest record (DCR) are efficient but have limited sensitivity to MIA risk. We introduce ReMIA (Relative Membership Inference Attack), a practical privacy metric that requires only two SDG training runs and additional data no larger than the original training set. Rather than predicting whether a record was in the training set, ReMIA generates two synthetic datasets from two source datasets and measures whether a classifier can identify which source a record came from. Experiments across multiple tabular datasets and SDGs show that ReMIA has a sensitivity comparable to state-of-the-art MIAs while being substantially more practical. We further observe that SDGs can achieve privacy-utility trade-offs that traditional noise-based anonymization methods do not match. Code is available at https://github.com/aindo-com/remia.

In physical systems, whenever a continuous symmetry is spontaneously broken, the system possesses excitations called Goldstone modes, which allow coherent information propagation over long distances and times. In this work, we study deep neural networks whose internal layers are equivariant under a continuous symmetry and may therefore support analogous Goldstone-like degrees of freedom. We demonstrate, both analytically and empirically, that these degrees of freedom enable coherent signal propagation across depth and recurrent iterations, providing a mechanism for stable information flow without relying on architectural stabilizers such as residual connections or normalization. In feedforward networks, this results in improved trainability and representational diversity across layers. In recurrent settings, we demonstrate the same mechanism is valuable for long-term memory by propagating information over recurrent iterations, thereby improving performance of RNNs and GRUs on long-sequence modeling tasks.

Cultural heritage institutions increasingly disseminate research and interpretive materials globally, but multilingual dissemination is constrained by limited budgets and staffing. In terminology-dense domains such as rock art, translation quality depends on accurate, consistent specialised terms, and small lexical errors can mislead non-specialists and reduce reuse. We compare three English MT setups for a Spanish academic rock art text, focusing on simple, operationally feasible interventions rather than complex model-side modifications: (1) DeepL as a strong NMT baseline, (2) Gemini-Simple (LLM with a basic prompt), and (3) Gemini-RAG (the same LLM with glossary-augmented prompting via term-pair retrieval). Using PEARMUT, we conduct a human evaluation via (i) multi-way Direct Assessment (0--100) and (ii) targeted terminology auditing with a restricted MQM taxonomy. Gemini-RAG yields the highest exact-match terminology accuracy (81.4\%), versus Gemini-Simple (69.1\%) and DeepL (64.4\%), while preserving overall quality (mean DA 85.3 Gemini-RAG vs. 85.2 Gemini-Simple), outperforming DeepL (80.3). These results show that glossary-augmented prompting is a low-overhead way to improve terminology control in cultural-heritage translation if institutions maintain minimal terminology resources and lightweight evaluation procedures.

Agentic AI systems are entering software engineering workflows, yet empirical evidence on how industrial organizations actually adopt them remains sparse. We present a qualitative interview study with sixteen practitioners across twelve companies of varying size and domain. This study characterizes the current agentic AI adoption state of these companies, employing a six-level maturity framework adapted from established AI-driven organizations. The findings reveal that seven companies operate at Level~1 (AI Assistants), four companies at Level~2 (AI Compensators), and only one in Level~3 (Multi-Agent Orchestration), with large and safety-regulated organizations among the most advanced adopters. The primary finding is a capability-deployment verification gap, four companies demonstrated higher-level experimental AI capabilities but cannot integrate them into production workflows because adequate output verification mechanisms are absent, leaving human-in-the-loop as the only trusted verification mechanism. This gap is shaped by four recurring barriers: context window of LLMs constraints especially when diverse knowledge aggregation is needed, under-performance on proprietary programming languages and protocols, non-determinism incompatible with qualification standards, and data confidentiality concerns. Two interdependent dimensions of this gap emerge from these findings (information asymmetry and qualification absence) framing a core open problem for industrial agentic integration.

Estimating an $N \times N$ quantum kernel from circuit fidelities requires $Θ(N^2 S)$ measurement shots, the dominant bottleneck for deployment on near-term hardware. Existing budget-saving methods (Nyström-QKE, ShoFaR, kernel-target alignment) sub-sample \emph{which} entries to measure but allocate shots \emph{uniformly} within their chosen subset, ignoring how much each entry drives the downstream classifier. We close this gap with two contributions. \textbf{First, a complete regime decomposition} for shot-budgeted quantum kernel learning: a principled menu of when each allocator wins. Our method, \emph{AQKA}, dominates the budget-limited regime ($B \lesssim 16 n_{\mathrm{pairs}}$) on sparse-sensitivity KRR, with the gap \emph{growing} from $+8$ to $+25$ pts over uniform as $N$ scales $225{\to}1000$ and reaching $+26$--$32$ pts on an \texttt{ibm\_pittsburgh} (156-qubit Heron) hardware kernel; Nyström-QKE wins at saturating budgets on planted-sparse via low-rank reconstruction; ShoFaR is competitive only at extreme low budgets. \textbf{Second, a closed-form pair-level acquisition theory}: $s_{ij}^{\star} \propto |g_{ij}|\sqrt{K_{ij}(1-K_{ij})}$ with explicit gradient $g_{ij}$ for KRR (Lemma~1, $|β_iα_j+β_jα_i|\sqrt{K_{ij}(1-K_{ij})}$) and SVM via the envelope theorem ($|η_i^*η_j^*|\sqrt{K_{ij}(1-K_{ij})}$); a \emph{corrected} sparsity-aware Cauchy--Schwarz rate $ρ\le 2m/N$ matching empirics (vs.\ the naive $m^2/N^2$); an explicit-constant plug-in regret bound (Theorem~2); and a tighter SVM ceiling $ρ^{\mathrm{SVM}} \le m_{\mathrm{sv}}^2/N^2$. We close with the first multi-seed live online adaptive shot allocation on quantum hardware: $+17.0 \pm 4.8$ pts at $N{=}20$ on \texttt{ibm\_aachen} ($3.5σ$, 5 seeds), with the advantage holding at $N{=}30$ at higher budget on \texttt{ibm\_berlin} ($+14.0 \pm 8.5$ pts, 5 seeds).

Autoresearch offers a flexible paradigm for automating scientific tasks, in which an AI agent proposes, implements, evaluates, and refines candidate solutions against a quantitative objective. Here, we use composition-based materials-property prediction to test whether such agents can perform a task beyond model selection and hyperparameter optimization: the design of input descriptors. We introduce Automat, an autoresearch framework where a coding agent based on a large language model generates composition-only descriptors for chemical compounds and evaluates them using a random forest workflow. The agent is restricted to information derivable from chemical formulas and iteratively proposes, implements, and tests chemically motivated descriptor strategies. We apply Automat, with OpenAI Codex using GPT-5.5 as the coding agent, to the prediction of experimental band gaps in inorganic materials and Curie temperatures in ferromagnetic compounds. In both tasks, Automat improves over fractional-composition, Magpie, and combined fractional-composition/Magpie baselines, while producing descriptor families that are chemically interpretable. These results provide a demonstration that autoresearch agents can generate competitive, task-specific materials descriptors without manual feature engineering during the run. They also reveal current limitations, including descriptor redundancy, sensitivity to greedy feature expansion, and the need for explicit complexity control, descriptor pruning, and more sophisticated search strategies.

A main barrier for the deployment of AI radiomic systems in clinical routine is their drop in performance under heterogeneous multicentre acquisition protocols. This work presents a performance-oriented framework for quantifying scan parameter sensitivity of radiomic AI models, while identifying clinically significant parameter regions associated with improved cross-dataset robustness. We formulate a mixed-effects framework for quantifying the influence that clinically relevant acquisition parameters have on models performance, while accounting for subject-level random effects. We have applied our framework to lung cancer diagnosis in CT scans using two independent multicentre datasets (a public database and own-collected data) and several SoA architectures. To evaluate across-database reproducibility, CT parameters have been adjusted using the data collected and tested on the public set. The optimal configuration selected is the current of the X-ray tube >= 200 mA, spiral pitch <= 1.5, slice thickness <= 1.25 mm, which balances diagnostic quality with low radiation dose. These configuration push metrics from 0.79+-0.04 sensitivity, 0.47+-0.10 specificity in low quality scans to 0.90+-0.10 sensitivity, 0.79 +- 0.13 specificity in high quality ones.

Dynamic systems in AI are often complex and heterogeneous, so that an internal specification is not accessible and verification techniques such as model checking are not applicable. Monitoring is in such cases an attractive alternative, as it evaluates desirable properties along traces generated by an unknown dynamic system. In this work, we consider anticipatory monitoring of linear-time properties enriched with an arbitrary SMT theory over finite traces (LTLfMT). Anticipatory monitoring in this setting is highly challenging, as the monitoring state depends on both the trace prefix seen so far and all its possible finite continuations. Under reasonable assumptions on the background theory, we present and formally prove the correctness of a novel foundational framework for monitoring properties in an expressive fragment of LTLfMT. The framework combines automata-theoretic methods to handle the temporal aspects of the logic, with automated reasoning techniques to address the first-order dimension. Moreover, we identify for the first time decidable fragments of this monitoring problem that are practically relevant as they combine linear arithmetic with uninterpreted functions, which covers e.g. data-aware business processes and dynamic systems operating over a read-only database. Feasibility is witnessed by a prototype implementation and preliminary evaluation.

Legal reasoning is not semantic similarity search. A court judgment encodes constrained symbolic reasoning: precedent propagation, procedural state transitions, and statute-bound inference. These are properties that vector-based retrieval-augmented generation (RAG) cannot faithfully represent. Hallucinated precedents, outdated statute citations, and unsupported reasoning chains remain persistent failure modes in LLM-based legal AI, with real consequences for access to justice in high-caseload jurisdictions such as India. This paper presents Falkor-IRAC, a graph-constrained generation framework for Indian legal AI that grounds generation in structured reasoning over an IRAC (Issue, Rule, Analysis, Conclusion) knowledge graph. Judgments from the Supreme Court and High Courts of India are ingested as IRAC node structures enriched with procedural state transitions, precedent relationships, and statutory references, stored in FalkorDB for low-latency agentic traversal. At inference time, LLM-generated answers are accepted only if a valid supporting path can be traced through the graph, a check performed by a falsifiability oracle called the Verifier Agent. The system also detects doctrinal conflicts as a first-class output rather than silently resolving them. Falkor-IRAC is evaluated using graph-native metrics: citation grounding accuracy, path validity rate, hallucinated precedent rate, and conflict detection rate. These metrics are argued to be more appropriate for legal reasoning evaluation than BLEU and ROUGE. On a proof-of-concept corpus of 51 Supreme Court judgments, the Verifier Agent correctly validated citations on completed queries and correctly rejected fabricated citations. Evaluation against vector-only RAG baselines is left for future work, as is GPU-accelerated inference to address current timeout rates on CPU hardware.

The multi-path Traveling Salesman Problem with stochastic travel costs arises in hybrid vehicle routing applications designed for Smart City and City Logistics, where multiple paths exist between each pair of locations. Travel times along these paths are typically affected by real-time traffic conditions and therefore modeled as stochastic. The objective of the problem is to determine a Hamiltonian tour that minimizes the expected total travel cost under uncertainty. In this work, we adopt a two-stage stochastic programming formulation. In the first stage, a predefined route specifying the sequence of locations to be visited is determined, while taking into consideration a second-stage recourse problem that selects the optimal path from the feasible set of alternative paths for each pair of locations, once real-time traffic conditions are realized. To reduce the computational burden imposed by the large number of scenarios required to capture travel time uncertainty, the innovation of this work is the integration of neural network-based surrogate models to approximate the expected value of the second-stage recourse problem. Different architectures and training strategies for the neural networks are proposed and analyzed, with performance evaluated in terms of computation time, solution quality, and generalization capability. Preliminary findings demonstrate the enhanced scalability and practical applicability of the approach for complex vehicle routing problems under uncertainty.

Post-Traumatic Stress Disorder (PTSD) is fundamentally a neuroplastic problem traumatic contact events encode over-reactive neural pathways through Hebbian long-term potentiation, producing hair-triggered amygdala-HPA stress cascades that fire before conscious awareness can intercept them. Existing therapeutic approaches, prolonged exposure, EMDR, cognitive behavioural therapy, operate predominantly downstream of the reactive cascade, teaching patients to tolerate or reframe distress after it has arisen. While clinically valuable, these suppression-based approaches do not produce the upstream pathway dissolution that constitutes lasting structural neural reorganisation. This paper proposes MindGap, a privacy-preserving on-device conversational AI framework that delivers structured neuroplastic rehabilitation for PTSD through the practice of dependent origination, a Buddhist psychological framework that identifies the precise moment between the pre-cognitive affective signal and the reactive elaboration that follows as the site of therapeutic intervention. MindGap guides patients through three progressive layers of observation at this feeling tone gap: noticing the bare affective signal before reactive elaboration, recognising it as self-arising rather than caused by the stimulus, and recognising the conditioned implicit belief beneath the feeling. Each layer corresponds to progressively deeper prefrontal regulatory engagement and progressively deeper long-term depression-mediated weakening of the reactive pathway, producing genuine upstream dissolution rather than downstream suppression. Running entirely on-device with no data egress, MindGap delivers daily calibrated exposure sessions through a fine-tuned lightweight large language model, making it deployable in sensitive clinical and military contexts where cloud-based solutions are not permitted.

Critical-data-size accounts of grokking suggest a natural post-threshold intuition: once training data is sufficient to identify the underlying rule, additional data should accelerate validation convergence. We show that this intuition can fail in a controlled structured-output task. In Needleman--Wunsch (NW) matrix generation, small Transformers reach high validation exact-match accuracy fastest at an intermediate dataset size, not at the largest one. Past this dataset-size sweet spot, generalization remains achievable but requires more gradient updates. Conversely, in the regime where partial validation competence first appears, larger datasets can require fewer updates to reach high training accuracy, suggesting that emerging rule structure can accelerate fitting beyond example-wise memorization. A multiplication baseline does not show the same post-threshold slowdown. These results separate the critical data size for the onset of generalization from the dataset size that optimizes update-based convergence, and identify structured-output tasks where learning the rule and completing exact-fitting can diverge.

With the rapid evolution of computer vision, vision-based methodologies for water level and river surface velocity estimation have reached significant maturity. Compared to traditional sensing, these techniques offer superior interpretability, automated data archiving, and enhanced system robustness. However, challenges such as environmental sensitivity, limited precision, and complex site calibration persist. This work proposes an integrated framework that synergizes state-of-the-art (SOTA) vision models with statistical modeling. By leveraging physical priors and robust filtering strategies, we improve the accuracy of water level detection and flow estimation. Code will be available at https://github.com/sunzx97/Vision_Based_Water_Level_and_Flow_Estimation.git

Deep learning methods based on backward stochastic differential equations (BSDEs) have emerged as competitive alternatives to physics-informed neural networks (PINNs) for solving high-dimensional partial differential equations (PDEs). By leveraging probabilistic representations, BSDE approaches can avoid the curse of dimensionality and often admit second-order-free training objectives that do not require explicit Hessian evaluations. It has recently been established that the commonly used Euler-Maruyama (EM) time discretization induces an intrinsic bias in BSDE training losses. While high-order schemes such as Heun can fully eliminate this bias, such schemes re-introduce second-order spatial derivatives and incur substantial computational overhead. In this work, we provide a principled analysis of EM-induced loss bias and propose an unbiased, second-order-free training framework that preserves the computational advantages of BSDE methods. Our code is available at https://github.com/seojaemin22/Un-EM-BSDE.

Class attribution maps (CAMs) provide local explanations for the decisions of convolutional neural networks. While widely used in practice, the evaluation of CAMs remains challenging due to the lack of ground-truth explanations, making it difficult to evaluate the soundness of existing metrics. Independently, most commonly used CAM methods produce low-resolution attribution maps, which limits their usefulness for detailed interpretability. To address the evaluation challenge, we introduce a synthetic dataset with ground-truth attributions that enables a rigorous comparison of CAM evaluation metrics. Using this dataset, we analyze existing metrics and propose ARCC, a new composite metric that more reliably identifies faithful explanations. To address the low resolution issue, we introduce RefineCAM, a method that produces high-resolution attribution maps by aggregating CAMs across multiple network layers. Our results show that RefineCAM consistently outperforms existing methods according to the proposed evaluation.

Large language models (LLMs) often fail to reason under temporal cutoffs: when prompted to answer from the standpoint of an earlier time, they exploit knowledge that became available only later. We study this failure through the lens of ex-ante reasoning, where a model must rely exclusively on information knowable before a cutoff. Through a systematic analysis of prompt-level interventions, we find that temporal leakage is highly sensitive to cutoff formulation and instruction placement: explicit cutoff statements outperform implicit historical framings, and prefix constraints reduce leakage more effectively than suffix constraints. These findings indicate that prompting can steer models into a temporal frame, but does not endow them with the ability to verify whether a response is temporally admissible. We further argue that supervised fine-tuning is insufficient, since ex-ante correctness is not an intrinsic property of an answer, but a relation between the answer and the cutoff. To address this gap, we propose TCFT, a Temporal Critique Fine-Tuning framework that trains models to acquire cutoff-aware temporal verification. Given a query, a cutoff, and a candidate response, TCFT teaches the model to identify post-cutoff leakage, explain temporal boundary violations, and judge temporal admissibility. Experiments with Qwen2.5-7B-Instruct and Qwen2.5-14B-Instruct show that TCFT consistently outperforms prompting and SFT baselines, reducing average leakage by 41.89 and 37.79 percentage points, respectively.

This paper introduces a multi-label visual emotion analysis benchmark dataset for comprehensively evaluating the ability of multimodal large language models (MLLMs) to predict the emotions evoked by images. Recent user studies report an unintuitive finding: humans may prefer the predictions of MLLMs over the labels in existing datasets. We argue that this phenomenon stems from the suboptimal annotation scheme used in existing datasets, where each annotator is shown a single candidate emotion for each image and judges whether it is evoked or not. This approach is clearly limited because a single image can evoke multiple emotions with varying intensities. As a result, evaluations based on these datasets may underestimate the capabilities of MLLMs, yet an appropriate benchmark for evaluating such models remains lacking. To address this issue, we introduce a new multi-label benchmark dataset for visual emotion analysis toward MLLMs evaluation. We hire $20$ annotators per image and ask them to select all emotions they feel from an image. Then, we aggregate the votes across all annotators, providing a more reliable and representative dataset labeled with a distribution of emotions. The resulting dataset contains $10,344$ images with $236,998$ valid votes across eight emotions. Based on this benchmark dataset, we evaluate several recent models, including Qwen3-VL, OpenAI's GPT, Gemini, and Claude. We assess model performance on both dominant emotion prediction and emotion distribution prediction. Our results demonstrate the progress achieved by recent MLLMs while also indicating that substantial room for improvement remains. Furthermore, our experiments with LLM-as-a-judge show that the method does not consistently improve MLLMs' performance, indicating its limitations for the subjective task of visual emotion analysis.

Migrating legacy C repositories to Rust promises stronger memory safety, but existing translators often work at the level of files or functions and miss architectural intent. We present RustPrint, a documentation-guided agentic framework for repository-level C-to-Rust migration. RustPrint first converts the source repository into architecture-aware documentation and treats it as a migration blueprint capturing module structure, data flow, APIs, and design rationale. Coding agents then use this blueprint to plan crates, implement modules, check compilability, reduce unsafe code, and iteratively refine the translated repository. RustPrint next compares documentation from the Rust output against the source documentation and uses mismatches as repair signals. It also translates and runs source test suites so runtime failures can guide targeted fixes. Experiments on eight real-world C repositories ranging from 11K to 84K LoC show that RustPrint compiles every target under both an open-weight (Kimi-K2-Instruct) and a closed-weight (GPT-5.4) backbone, while prior LLM-based translators (Self-Repair, EvoC2Rust) fail repository-wide. With the open-weight Kimi-K2-Instruct backbone, RustPrint exceeds an agentic Claude Code baseline on feature preservation (93.26% vs. 52.52%) and on cross-evaluation test pass rate (95.17% vs. 79.85%). These results suggest that documentation-guided coordination is a useful direction for scalable codebase migration.

Forecasting multivariate hidden Markov processes is challenging due to nonlinear and nonstationary observations, latent state transitions, and cross-sequence dependencies. While deep learning methods achieve strong predictive accuracy, they typically lack explicit state modeling, whereas Hidden Markov Models (HMMs) provide interpretable latent states but struggle with complex nonlinear emissions and scalability. To address these limitations, we propose DRL-STAF, a Deep Reinforcement Learning based STate-Aware Forecasting framework that jointly predicts next-step observations and estimates the corresponding hidden states for complex multivariate hidden Markov processes. Specifically, DRL-STAF models complex nonlinear emissions using deep neural networks and estimates discrete hidden states using reinforcement learning, reducing the reliance on predefined transition structures and enabling flexible adaptation to diverse temporal dynamics. In particular, DRL-STAF mitigates the state-space explosion encountered by typical multivariate HMM-based methods. Extensive experiments demonstrate that DRL-STAF outperforms HMM variants, standalone deep learning models, and existing DL-HMM hybrids in most cases, while also providing reliable hidden-state estimates.

We introduce Action-Inspired Generative Models (AGMs), a dual-network generative framework motivated by the observation that existing bridge-matching methods assign uniform regression weight to every stochastic transition in the transport landscape, regardless of whether a given bridge sample lies along a structurally coherent trajectory or a degenerate one. We address this by introducing a lightweight learned scalar potential $V_φ$ that scores bridge samples online and modulates the drift objective via importance weights derived through a stop-gradient barrier -- preventing adversarial feedback between the two networks whilst preserving $V_φ$'s guiding signal. Crucially, $V_φ$ comprises only $\sim$1.4% of the primary drift network's parameter count, adds no overhead to the inference graph, and requires no iterative half-bridge fitting or auxiliary stochastic differential equation (SDE) solvers: it is a plug-and-play enhancement to any bridge-matching training loop. At inference, $V_φ$ is discarded entirely, leaving standard Euler-Maruyama integration of the exponential moving average (EMA) drift. We demonstrate that selectively penalising uninformative transport paths through the learned potential yields consistent improvements in generation quality across fidelity and coverage metrics.

A common critique of neural combinatorial-optimization solvers is that they are less energy-efficient than CPU metaheuristics, given the operational energy cost of training them on GPUs. This paper examines the inferential step from "training is expensive" to "neural solvers are net-inefficient", which is where the critique actually goes wrong. Training the network costs a large fixed amount of GPU energy; running the metaheuristic costs a small amount of CPU energy on every instance, repeated as long as the solver is deployed. The two are not commensurable until a deployment volume is fixed. We define the Amortized Efficiency Threshold (AET) as the deployment volume above which a neural solver breaks even with a heuristic baseline in total energy or carbon, under an explicit constraint on solution quality. We show that the cumulative-energy ratio between the two solvers tends to a constant strictly below one whenever the network wins per-instance, and that this limit does not depend on how the training cost was measured. An embodied-carbon term amortizes hardware fabrication symmetrically on both sides. We instantiate the framework on the Multi-Task VRP (MTVRP) environment at n=20 customers across 19 problem variants and five training seeds, with HGS via PyVRP as the heuristic baseline. The measured crossover sits near $1.58 \times 10^5$ deployed instances; the per-instance ratio is 0.41, reflecting the moderate size of the instances tested. The contribution is the framework, the open instrumentation, and the measurement protocol; structural convergence of the ratio at larger problem sizes is left to future empirical work.

Large vision-language models (LVLMs) often hallucinate when language priors dominate weak or ambiguous visual evidence. Existing contrastive decoding methods mitigate this problem by comparing predictions from the original image with those from externally perturbed visual inputs, but such references can introduce off-manifold artifacts and require costly extra forward passes. We propose SIRA, a training-free internal contrastive decoding framework that constructs a counterfactual reference inside the same LVLM by exploiting the staged information flow of multimodal transformers. Instead of removing visual information from the input, SIRA first lets image and text tokens interact through a shared prefix, forming an aligned multimodal state that preserves prompt interpretation, decoding history, positional structure, and early visual grounding. It then forks a counterfactual branch in later transformer layers, where attention to image-token positions is masked. This branch retains the shared multimodal context but lacks continued access to fine-grained visual evidence, yielding a language-prior-dominated internal reference for token-level contrast. During decoding, SIRA suppresses tokens that remain strong without late visual access and favors predictions whose advantage depends on the full visual pathway. Experiments on POPE, CHAIR, and AMBER with Qwen2.5-VL and LLaVA-v1.5 show that SIRA consistently reduces hallucinations while preserving descriptive coverage and incurring lower overhead than two-pass contrastive decoding. SIRA requires no training, external verifier, or perturbed input, and applies to open-weight LVLMs with white-box inference access.

Multi-run chain-of-thought reasoning is usually collapsed to final-answer aggregates, which discard howsampled trajectories share, split, and rejoin through intermediate computation. We propose SliceGraph, a post-hoc problem-model-cell graph built by mutual-kNN over sparse activation-key Jaccard similarity between CoT slices, and treat it as a measurement object for process geometry rather than as a decoding program. Across sampled CoT ensembles from three primary 4B/8B models on math and science benchmarks, blinded annotation supports SliceGraph biconnected components as shared reasoning-state units and process families as within-family strategy-coherent route units. In 85.5% of 954 problem-model cells, correct CoTs sharing the same normalized answer split into multiple process families; among cells with at least two such runs, 76.6% of run pairs are cross-family on average. We call such same-answer, family-divergent correct trajectories process isomers. A label-seeded reward field provides a separate value-landscape layer: success-associated regions often split into disconnected high-value cores, and route families specialize over these core footprints rather than merely duplicating one another. A typed-state transition analysis further shows that process families navigate the same atlas with distinct transition kernels under matched null controls. Representation ablations, a cross-architecture replication, and two cross-scale replications support the robustness of the route-family scaffold, showing that final-answer aggregation overlooks this structured multi-route process geometry.

AI-enabled features built on LLMs and agentic workflows are difficult to test, debug, and reproduce, especially for product-focused software engineers without a machine learning background. We present the AI Toolkit plugin for JetBrains IDEs, which brings tracing and evaluation directly into the Run/Debug loop. A mixed methods study with practitioners presents three consistent needs: (1) make evaluation regular and repeatable, (2) expose traces at the moment of execution, and (3) minimize setup and context switching. Guided by these needs, the AI Toolkit introduces an IDE-native workflow: run-triggered trace capture; immediate, hierarchical inspection; one-click "Add to Dataset" from traces; and unit-test-like evaluations with pluggable metrics. The first release in PyCharm shows promising early signals - strong conversion when promoted at Run, sustained usage among those who capture traces, and low churn - suggesting that IDE-native observability lowers activation energy and helps developers adopt disciplined practices. We detail the design and implementation of the AI Agents Debugger and AI Evaluation, report initial adoption telemetry, and outline next steps to broaden framework coverage and scale evaluations. Together, these results indicate that integrating AI observability and evaluation into everyday IDE workflows can make modern AI development accessible to non-ML specialists while preserving software-engineering practices.

Accurate image segmentation remains challenging, particularly in generating sharp, confident boundaries. While modern architectures have advanced the field, many of them still rely on standard loss functions like Cross-Entropy and Dice, which often neglect the discriminative structure of learned features, leading to inaccurate boundaries. This work introduces Deep Discriminant Analysis (DDA), a differentiable, architecture-agnostic loss function that embeds classical discriminant principles for network training. DDA explicitly maximizes between-class variance while minimizing within-class one, promoting compact and separable feature distributions without increasing inference cost. Evaluations on the DIS5K benchmark demonstrate that DDA consistently improves segmentation accuracy, boundary sharpness, and model confidence across various architectures. Our results show that integrating discriminant analysis offers a simple, effective path for building more robust segmentation models.

Model providers increasingly release open weights or allow users to fine-tune foundation models through APIs. Although these models are safety-aligned before release, their safeguards can often be removed by fine-tuning on harmful data. Recent defenses aim to make models robust to such malicious fine-tuning, but they are largely evaluated only against fixed attacks that do not account for the defense. We show that these robustness claims are incomplete. Surveying 15 recent defenses, we identify several defense mechanisms and show that they share a single weakness: they obscure or misdirect the path to harmful behavior without removing the behavior itself. We then develop a unified adaptive attack that breaks defenses across all defense mechanisms. Our results show that current approaches do not provide robust security; they mainly stop the attacks they were designed against. We hope that our unified adaptive adversary for this domain will help future researchers and practitioners stress-test new defenses before deployment.

This position paper argues that effective tutoring requires corrective friction: surfacing misconceptions and challenging them supportively to drive conceptual change. Yet preference-aligned LLMs can trade epistemic rigor for agreeableness. We identify a Reasoning-Sycophancy Paradox: models that resist context-switch frame attacks can still capitulate under social-epistemic pressure, especially authority ("my notes say I'm right") and social-affective face-saving ("please don't tell me I'm wrong"). We introduce EduFrameTrap, a tutoring benchmark across math, physics, economics, chemistry, biology, and computer science that varies student confidence and pressure (context-switch, authority, social-affective). Across two frontier LLMs, context-switch failures are comparatively lower for GPT-5.2, while authority and social pressure more often trigger epistemic retreat. In contrast, Claude shows substantial context-switch fragility in this run. Because these failures are hard to judge automatically, we report two-judge disagreement as a reliability signal. We argue benchmarks should measure social-epistemic courage, i.e., supportive but corrective tutoring, and treat kind-but-correct behavior as a safety requirement.

Scientific progress depends on sequences of enabling contributions, yet existing AI4Science benchmarks largely focus on citation prediction, literature retrieval, or idea generation rather than the dependencies that make progress possible. In this paper, we introduce discovery pathway forecasting: given a target scientific contribution and the prior literature available at a specified time, the task is to (1) identify the enabling contributions required to realize it and (2) ground each in prior work when such prior work exists. We present SciPaths, a benchmark of 262 expert-annotated gold pathways and 2,444 silver pathways constructed from machine learning and natural language processing papers, where each pathway records enabling contributions, roles, rationales, and prior-work groundings or unmapped decisions. Evaluating frontier and open-weight language models, we find that the best model reaches only 0.189 F1 under strict semantic matching, with core methodological dependencies hardest to recover. Prior-work grounding improves substantially when gold enabling contributions are provided, showing that decomposition quality is a major bottleneck for end-to-end pathway recovery. SciPaths therefore shifts evaluation toward a missing capability in scientific forecasting: reasoning backward from a target contribution to the enabling scientific building blocks and prior-work dependencies that make it feasible.

We establish novel structural and statistical results for entropy-regularized min-max inverse reinforcement learning (Min-Max-IRL) with linear reward classes in finite-horizon MDPs with Borel state and action spaces. On the structural side, we show that maximum likelihood estimation (MLE) and Min-Max-IRL are equivalent at the population level, and at the empirical level under deterministic dynamics. On the statistical side, exploiting pseudo-self-concordance of the Min-Max-IRL loss, we prove that both the trajectory-level KL divergence and the squared parameter error in the Hessian norm decay at the fast rate $\mathcal{O}(n^{-1})$, where $n$ is the number of expert trajectories. Our guarantees apply under misspecification and require no exploration assumptions. We further extend reward-identifiability results to general Borel spaces and derive novel results on the derivatives of the soft-optimal value function with respect to reward parameters.

Extending the context window of large language models typically requires training on sequences at the target length, incurring quadratic memory and computational costs that make long-context adaptation expensive and difficult to reproduce. We propose EndPrompt, a method that achieves effective context extension using only short training sequences. The core insight is that exposing a model to long-range relative positional distances does not require constructing full-length inputs: we preserve the original short context as an intact first segment and append a brief terminal prompt as a second segment, assigning it positional indices near the target context length. This two-segment construction introduces both local and long-range relative distances within a short physical sequence while maintaining the semantic continuity of the training text--a property absent in chunk-based simulation approaches that split contiguous context. We provide a theoretical analysis grounded in Rotary Position Embedding and the Bernstein inequality, showing that position interpolation induces a rigorous smoothness constraint over the attention function, with shared Transformer parameters further suppressing unstable extrapolation to unobserved intermediate distances. Applied to LLaMA-family models extending the context window from 8K to 64K, EndPrompt achieves an average RULER score of 76.03 and the highest average on LongBench, surpassing LCEG (72.24), LongLoRA (72.95), and full-length fine-tuning (69.23) while requiring substantially less computation. These results demonstrate that long-context generalization can be induced from sparse positional supervision, challenging the prevailing assumption that dense long-sequence training is necessary for reliable context-window extension. The code is available at https://github.com/clx1415926/EndPrompt.

Recursive learning -- where models are trained on data generated by previous versions of themselves -- is increasingly common in large language models, autonomous agents, and self-supervised systems. However, standard performance metrics (loss, perplexity, accuracy) often fail to detect internal degradation before it becomes irreversible. Here we identify a phenomenon we call silent collapse: under broad recursive conditions, model internal distributions -- predictive entropy, representational diversity, and tail coverage -- progressively contract even as conventional metrics appear stable or improving. We discover that silent collapse is not abrupt. Its onset is reliably preceded by three trajectory-level precursors: (1) contraction of anchor entropy, (2) freezing of representation drift, and (3) erosion of tail coverage. These signals manifest multiple generations before any degradation in standard validation metrics, enabling early warning. Based on these precursors, we propose the MTR (Monitor--Trust--Regulator) framework, a lightweight metacognitive loop that monitors trajectory statistics, estimates a slow-timescale trust variable, and adaptively modulates the effective learning intensity. MTR provides early warning and actively prevents silent collapse without requiring access to pristine real data -- a critical advantage when original data is unavailable, contaminated, or private.

Extensive research has highlighted the severe threats posed by backdoor attacks to deep reinforcement learning (DRL). However, prior studies primarily focus on vanilla scenarios, while plasticity interventions have emerged as indispensable built-in components of modern DRL agents. Despite their effectiveness in mitigating plasticity loss, the impact of these interventions on DRL backdoor vulnerabilities remains underexplored, and this lack of systematic investigation poses risks in practical DRL deployments. To bridge this gap, we empirically study 14,664 cases integrating representative interventions and attack scenarios. We find that only one intervention (i.e., SAM) exacerbates backdoor threats, while other interventions mitigate them. Pathological analysis identifies that the exacerbation is attributed to backdoor gradient amplification, while the mitigation stems from activation pathway disruption and representation space compression. From these findings, we derive two novel insights: (1) a conceptual framework SCC for robust backdoor injection that deconstructs the mechanistic interplay between interventions and backdoors in DRL, and (2) abnormal loss landscape sharpness as a key indicator for DRL backdoor detection.

Implicit solvent models are widely used to decrease the number of solvent degrees of freedom and enable the calculation of solvation energetics without water molecules. However, its accuracy often falls short compared to explicit models. Recent advancements in neural potentials have shown promise in drug discovery, but transferability remains a persistent challenge. Here, we introduce the Protein Hydration Neural Network (PHNN), an implicit solvent model that extends analytical continuum solvation by learning transferable corrections to model parameters instead of applying post hoc adjustments to final energies. The model is explicitly designed to maximize data efficiency by leveraging physical priors embedded in the data. We demonstrate that PHNN improves accuracy relative to traditional analytical methods and maintains predictive accuracy on out-of-domain protein systems.

Visual RAG has offered an alternative to traditional RAG. It treats documents as images and uses vision encoders to obtain vision patch tokens. However, hundreds of patch tokens per document create retrieval and storage challenges in a vector database. Practical deployment requires aggregating them into a single vector. This raises a critical question: does single-vector aggregation lose key information in financial documents? We develop a diagnostic benchmark using financial documents where changes in single digits can lead to significant semantic shifts. Our experiments show that single-vector aggregation collapses different documents with almost identical vectors. Metrics show that the patch level detects semantic changes, and confirm that aggregation obscures these details. We identify global texture dominance as the root cause. Our findings are consistent across model scales, retrieval-optimized embeddings, and multiple mitigation strategies, highlighting significant risks for single-vector visual document retrieval in financial applications.

Partial Dependence Plots (PDPs) visualize how changes in a single feature affect the average model prediction. They are widely used in practice to interpret decision tree ensembles and other machine learning models. Joint-PDPs extend this idea to pairs of features, revealing their combined effect. Partial Dependence Interaction Values (PDIVs) measure feature interactions. The Any-Order-PDIVs task computes these interactions for every feature subset across all rows of the dataset. We introduce Woodelf++, a unified and efficient approach for computing all these useful explainability tools on decision tree ensembles, building on Woodelf, an algorithm for efficient SHAP computation. By deriving suitable metrics over pseudo-Boolean functions, Woodelf++ can compute PDPs (exact and approximate), Joint-PDPs, and Any-Order-PDIVs in a unified framework. Our method delivers substantial complexity improvements over the state of the art, including an exponential gain for Any-Order-PDIVs. Additionally, we introduce and efficiently compute Full PDPs, which leverage the model's split thresholds to faithfully capture its behavior across all possible feature values. Woodelf++ is implemented in pure Python and supports GPU acceleration. On a dataset with 400,000 rows, Woodelf++ computes PDP and Joint-PDP up to 6x faster than the state of the art and up to five orders of magnitude faster than scikit-learn. For Any-Order-PDIVs, the gap is even larger: Woodelf++ computes all interaction values in 5 minutes, while the state of the art is estimated to require over 1,000,000 years.

Observing touch on another's body can elicit corresponding tactile sensations in the observer, a phenomenon termed mirror touch that supports empathy and social perception. This visuo-tactile resonance is thought to rely on structural correspondence between visual and somatosensory cortices, yet robotic systems lack computational frameworks that instantiate this principle. Here we demonstrate that cortical correspondence can be operationalized to endow robots with mirror touch. We introduce Mirror Touch Net, which imposes semantic, distributional and geometric alignment between visual and tactile representations through multi-level constraints, enabling prediction of millimetre-scale tactile signals across 1,140 taxels on a robotic hand from RGB images. Manifold analysis reveals that these constraints reshape visual representations into geometry consistent with the tactile manifold, reducing the complexity of cross-modal mapping. Extending this alignment framework to cross-domain observations of human hands enables tactile prediction and reflexive responses to observed human touch. Our results link a neural principle of visuo-tactile resonance to robotic perception, providing an explainable route towards anticipatory touch and empathic human-robot interaction. Code is available at https://github.com/fun0515/Mirror-Touch-Net.

Large Language Diffusion Models (LLDMs) are emerging as an alternative to autoregressive models, offering faster inference through higher parallelism. Similar to autoregressive LLMs, they remain prone to hallucinations, making reliable uncertainty quantification (UQ) crucial for safe deployment. However, existing UQ methods are fundamentally misaligned with this new paradigm: they assume autoregressive factorization or use expensive repeated sampling, negating the efficiency of LLDMs. In this work, we present the first systematic study of UQ for LLDMs and propose lightweight, zero-shot uncertainty signals derived from the iterative denoising process, leveraging intermediate generations, token remasking dynamics, and denoising complexity. We further adapt a state-of-the-art UQ method to LLDMs by combining masked diffusion likelihoods with trajectory-based semantic dissimilarity. We prove that expected trajectory dissimilarity lower bounds the masked diffusion training objective, which motivates its usage as an uncertainty score. Comprehensive experiments across three tasks, eight datasets, and two models show that our method achieves a great cost-performance trade-off: it approaches the strongest sampling-based baselines while incurring up to 100x lower computational overhead. Our work demonstrates that LLDMs can deliver both fast inference and reliable hallucination detection simultaneously.

Context. Behaviour-Driven Development (BDD) software test suites accumulate duplicated step subsequences. Three published refactoring patterns are available (within-file Background, within-repo reusable-scenario invocation, cross-organisational shared higher-level step), but no prior work automates which recurring subsequences are worth extracting or which mechanism applies. Objective. Rank recurring step subsequences ("slices") by refactoring suitability (extraction-worthy), pre-map each to one of the three patterns, and quantify prevalence across the public BDD ecosystem. Method. Every contiguous L-step window (L in [2, 18]) in a 339-repository / 276-upstream-owner Gherkin corpus is keyed by paraphrase-robust cluster identifiers and counted under three scopes. Sentence-BERT (SBERT) / Uniform Manifold Approximation and Projection (UMAP) / Hierarchical Density-Based Clustering (HDBSCAN) recovers paraphrase-equivalent slices. Three authors label a stratified 200-slice pool against a written rubric. An eXtreme Gradient Boosting (XGBoost) extraction-worthy classifier trained under 5-fold cross-validation is compared with a tuned rule baseline and two open-weight Large Language Model (LLM) judges. Results. The miner produces 5,382,249 slices collapsing to 692,020 recurring patterns. Three-author Fleiss' kappa = 0.56 (extraction-worthy) and 0.79 (mechanism). The classifier reaches out-of-fold F1 = 0.891 (95% CI [0.852, 0.927]), outperforming both the rule baseline (F1 = 0.836, p = 0.017) and the better LLM judge (F1 = 0.728, p < 1e-4). 75.0%, 59.5%, and 11.7% of scenarios carry a within-file Background, within-repo reusable-scenario, or cross-organisational shared-step candidate. Conclusion. Paraphrase-robust subscenario discovery yields a corpus-wide census of BDD refactoring opportunities; pipeline, classifier predictions, labelled pool, and rubric are released under Apache-2.0.

We propose a simple mechanism by which scaling laws emerge from feature learning in multi-layer networks. We study a high-dimensional hierarchical target that is a globally high-degree function, but that can be represented by a combination of latent compositional features whose weights decrease as a power law. We show that a layer-wise spectral algorithm adapted to this compositional structure achieves improved scaling relative to shallow, non-adaptive methods, and recovers the latent directions sequentially: strong features become detectable at small sample sizes, while weaker features require more data. We prove sharp feature-wise recovery thresholds and show that aggregating these transitions yields an explicit power-law decay of the prediction error. Technically, the analysis relies on random matrix methods and a resolvent-based perturbation argument, which gives matching upper and lower bounds for individual eigenvector recovery beyond what standard gap-based perturbation bounds provide. Numerical experiments confirm the predicted sequential recovery, finite-size smoothing of the thresholds, and separation from non-hierarchical kernel baselines. Together, these results show how smooth scaling laws can emerge from a cascade of sharp feature-learning transitions.

Automated code documentation is essential for modern software development, providing the contextual grounding that both human developers and coding agents rely on to navigate large codebases. Existing repository-level approaches process components independently, causing redundant retrieval and conflicting descriptions across documents while producing outputs that lack hierarchical structure. Therefore, we propose MemDocAgent, a long-horizon agentic framework that generates documentation within a single, integrated context spanning the entire repository. It combines two components: (i) Dependency-Aware Traversal Guiding that predetermines a traversal order respecting dependency and granularity hierarchies; (ii) Memory-Guided Agentic Interaction, in which the agent interacts with RepoMemory, a shared memory accumulating prior work traces through read, write, and verify operations. Through an in-depth multi-criteria evaluation, MemDocAgent achieves the best performance over both open and closed-source baselines and demonstrates practical applicability in real software development workflows.

Prompt engineering is crucial for effective interaction with generative artificial intelligence systems, yet existing optimisation methods often operate over an unstructured and vast prompt space, leading to high computational costs and potential distortions of the original intent. We introduce Prompt Segmentation and Annotation Optimisation (PSAO), a structured prompt optimisation framework designed to improve prompt optimisation controllability and efficiency. PSAO decomposes a prompt into interpretable segments (e.g., sentences) and augments each with human-readable annotations (e.g., {not important}, {important}, {very important}). These annotations guide large language models (LLMs) in allocating focus and clarifying confusion during response generation. We formally define the segmentations and annotations and demonstrate that optimised segment-level annotations can lead to improved LLM responses, with the original prompt retained as a candidate in the optimisation space to prevent performance degradation. Empirical evaluations indicate that PSAO benefits from annotations in terms of improved reasoning accuracy and self-consistency. However, developing efficient methods for identifying optimal segmentations and annotations remains challenging and is reserved for future investigation. This work is intended as a proof of concept, demonstrating the feasibility and potential of segment-level annotation optimisation.

PyCSP$^3$ provides a productive way to build constraint models for solving combinatorial constrained problems and export them to XCSP$^3$, preserving a complete separation between modeling and solving. However, it lacks native support for scheduling abstractions such as interval variables, sequence variables, and resource functions. As a result, scheduling models must be encoded with low-level integer variables and manual channeling constraints, even though PyCSP$^3$ already provides global constraints like NoOverlap and Cumulative on integer arrays. We present PyCSP$^3$ Scheduling, a library that adds scheduling abstractions to PyCSP$^3$ through 53 dedicated constraints and 27 expressions, and compiles them down to standard PyCSP$^3$/XCSP$^3$ constraints, maintaining the modeling/solving separation that underpins the PyCSP$^3$ ecosystem. On 261 paired instances across 17 model families (5 runs each), both formulations produce identical objectives on all 72 doubly-proved optimal pairs and nearly half of the families (8/17) remain structurally unchanged after compilation; however, runtime performance diverges across families, with clear gains on some (up to 5.8x) and regressions on others due to the overhead of compilation decompositions. Code and benchmarks are available at: https://github.com/sohaibafifi/pycsp3-scheduling

Agentic reinforcement learning trains large language models using multi-turn trajectories that interleave long reasoning traces with short environment-facing actions. Common policy-gradient methods, such as PPO and GRPO, treat each token in a trajectory equally, leading to uniform credit assignment. In this paper, we critically demonstrate that such uniform credit assignment largely misallocates token-level training signals. From an energy-based modeling perspective, we show that token-level training signals, quantified by their correlations with reward variance of different rollouts sampled from a given prompt, concentrate sharply on action tokens rather than reasoning tokens, even though action tokens account for only a small fraction of the trajectory. We refer to this phenomenon as the Action Bottleneck. Motivated by this observation, we propose an embarrassingly simple token reweighting approach, ActFocus, that downweights gradients on reasoning tokens, along with an additional energy-based redistribution mechanism that further increases the weights on action tokens with higher uncertainty. Across four environments and different model sizes, ActFocus consistently outperforms PPO and GRPO, yielding final-step gains of up to 65.2 and 63.7 percentage points, respectively, without any additional runtime or memory cost.

Symmetrical Reality (SR) is emerging as a future trend for human-agent coexistence, placing higher demands on agents to acquire human-like intelligence. It calls for richer and more diverse human guidance. We introduce a three-stage demonstration paradigm integrating multimodal demonstration signals. Building on this paradigm, we developed TeachAnything, a cloud-based, crowdsourcing-oriented demonstration platform with physics simulation capable of collecting diverse demonstration data across varied scenes, tasks, and embodiments. By unifying virtual and physical interactions through both methodological design and physics simulation, the system serves as a practical foundation for developing embodied agents aligned with Symmetrical Reality.

Current methods for creating drum loop audio in digital music production, such as using one-shot samples or resampling, often demand non-trivial efforts of creators. While recent generative models achieve high fidelity and adhere to text, they lack the specific control needed for such a task. Existing symbolic-to-audio research often focuses on single, tonal instruments, leaving the challenge of polyphonic, percussive drum synthesis unaddressed. We address this gap by introducing ``Break-the-Beat!,'' a model capable of rendering a drum MIDI with the timbre of a reference audio. It is built by fine-tuning a pre-trained text-to-audio model with our proposed content encoder and a effective hybrid conditioning mechanism. To enable this, we construct a new dataset of paired target-reference drum audio from existing drum audio datasets. Experiments demonstrate that our model generates high-quality drum audio that follows high-resolution drum MIDI, achieving strong performance across metrics of audio quality, rhythmic alignment, and beat continuity. This offer producers a new, controllable tool for creative production. Demo page: https://ik4sumii.github.io/break-the-beat/

Prompt engineering has become central to eliciting the capabilities of large language models (LLMs). At its core lies prompt selection -- efficiently identifying the most effective prompts. However, most prior investigations overlook a key challenge: the inherently multi-faceted nature of prompt performance, which cannot be captured by a single metric. To fill this gap, we study the multi-objective prompt selection problem under two practical settings: Pareto prompt set recovery and best feasible prompt identification. Casting the problem into the pure-exploration bandits framework, we adapt provably efficient algorithms from multi-objective bandits and further introduce a novel design for best feasible arm identification in structured bandits, with theoretical guarantees on the identification error in the linear case. Extensive experiments across multiple LLMs show that the bandit-based approaches yield significant improvements over baselines, establishing a principled and efficient framework for multi-objective prompt optimization.

Instance normalization (IN) is widely used in non-stationary multivariate time series forecasting to reduce distribution shifts and highlight common patterns across samples. However, IN can over-smooth instance-specific structural information that is essential for modeling temporal and cross-channel heterogeneity. While prior methods further suppress distribution discrepancies or attempt to recover temporal specific dependencies, they often ignore a central tension: how to adaptively model common and instance-specific dependency based on each instance's non-stationary structures. To address this dilemma, we propose SeesawNet, a unified architecture that dynamically balances common and instance-specific dependency modeling in both temporal and channel dimensions. At its core is Adaptive Stationary-Nonstationary Attention (ASNA), which captures common dependencies from normalized sequences and specific dependencies from raw sequences, and adaptively fuses them according to instance-level non-stationarity. Built upon ASNA, SeesawNet alternates dedicated temporal and channel relationship modeling to jointly capture long-range and cross-variable dependencies. Extensive experiments on multiple real-world benchmarks demonstrate that SeesawNet consistently outperforms state-of-the-art methods.

Artificial intelligence in high-stakes tabular domains cannot be evaluated by predictive performance alone, yet current practice still assesses explainability, fairness, robustness, privacy, and sustainability mostly in isolation. We propose the Model Integrity and Responsibility Assessment Index (MIRAI), a unified evaluation framework that measures tabular models across these five dimensions under a controlled comparison setting and aggregates them into a single score. MIRAI combines established metrics through normalized and direction-aligned dimension scores, which enables direct comparison across models with different architectural and computational profiles. Experiments on healthcare, financial, and socioeconomic datasets show that higher predictive performance does not necessarily imply better overall integrity and responsibility. In several cases, simpler models achieve a stronger cross-dimensional balance than more complex deep tabular architectures. MIRAI provides a compact and practical basis for responsible model selection in regulated settings.

Recent advances in neural operators have made partial differential equation (PDE) surrogate modeling increasingly scalable and transferable through large-scale pretraining and in-context adaptation. However, after a shared operator is fine-tuned to multiple regimes within a continuous physical family, it remains unclear whether the resulting weight-space updates merely form isolated regime experts or reveal reusable physical structure. Starting from a shared family anchor, we fine-tune low- and high-regime endpoint experts and show that their updates can be separated into a family-shared adaptation and a direction aligned with the underlying physical parameter. This separation reinterprets endpoint experts as finite-difference probes of a local physical direction in weight space, explaining why static averaging can interpolate between regimes but attenuates endpoint-specific physics. Building on this perspective, we propose Calibration-Conditioned Merge (CCM), a post-hoc coordinate readout method for composing neural PDE experts along this physical direction. Given physical metadata, a calibrated coordinate mapping, or a short observed rollout prefix, CCM infers the target composition coordinate and deploys a single merged checkpoint for the remaining rollout. We evaluate CCM on the reaction--diffusion system, viscosity-parameterized two-dimensional Navier--Stokes equations, and radial dam-break dynamics. Across these benchmarks, CCM achieves its strongest gains in extrapolative regimes, reducing out-of-distribution rollout error relative to the family anchor by 54.2%, 42.8%, and 13.8%, respectively. Further experiments across FNO scales, a DPOT-style backbone, and ablations confirm that endpoint fine-tuning is not arbitrary checkpoint drift, but reveals a calibratable physical direction for training-free transfer across PDE regimes.

Large language models are often described as sycophantic, in the sense that they appear to flatter users or mirror their beliefs. We argue that this label is conceptually misleading: sycophancy implies motives and strategic intent, which LLMs do not possess. Their behaviour is better understood as complacency, a structural tendency to agree with user input because training data, reward signals and design favour agreement and reinforcement over correction. We argue that this distinction matters. Whether developers act sycophantically or not, models themselves never are sycophants; they can only be made more or less complacent. This reframing locates agency in developers and institutions, not in the model. Because complacent models reinforce users' prior beliefs, we argue that AI literacy educational approaches should particularly focus on strategies to counter confirmation bias.

Inpatient medication recommendation requires clinicians to repeatedly select specific medications, doses, and routes as a patient's condition evolves. Existing benchmarks formulate this task as admission-level prediction over coarse drug codes with multi-hot diagnostic and procedure code inputs, failing to capture the per-timepoint, information-rich nature of real prescribing. We propose RxEval, a prescription-level benchmark that evaluates LLM prescribing capability by multiple-choice questions: each question presents a detailed patient profile and time-ordered clinical trajectory, requiring selection of specific medication-dose-route triples from real prescriptions and patient-specific distractors generated via reasoning-chain perturbation. RxEval comprises 1,547 questions spanning 584 patients, 18 diagnostic categories, and 969 unique medications. Evaluation of 16 LLMs shows that RxEval is both challenging and discriminative: F1 ranges from 45.18 to 77.10 across models, and the best Exact Match is only 46.10%. Error analysis reveals that even frontier models may overlook stated patient information and fail to derive clinical conclusions.

A skilled live-commerce host is not merely a narrator, but a sales agent who converts viewer curiosity into purchase intent through expert product knowledge, emotionally intelligent response tactics, and entertainment that serves as a vehicle for product exposure. Yet no existing AI system replicates this: conversational recommenders treat recommendation as a terminal act, while general-purpose LLMs hallucinate product claims and default to generic promotional templates that fail to engage or persuade. We present VerbalValue, a sales-conversion-oriented virtual host that turns exceptional verbal ability into real commercial value, built on three contributions. First, we construct a domain knowledge base of product specifications and a curated sales terminology lexicon that anchor product-related responses in verified expertise. Second, we collect and annotate 1,475 live-commerce interactions spanning diverse viewer intents. Third, we fine-tune a large language model on this data to deliver empathetic, commercially oriented responses, adapting to viewer intent through empathetic amplification, evidence-backed rebuttal, and humor-mediated deflection. Experiments against GPT-5.4, Claude Sonnet 4.6, Gemini 3.1 Pro, and other baselines demonstrate gains of 23% on informativeness and 18% on factual correctness, with consistent advantages in tactfulness and viewer engagement.

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as an effective paradigm for improving the reasoning capabilities of large language models. However, RLVR training is often hindered by sparse binary rewards and weak credit assignment, resulting in ambiguous optimization signals and underutilization of the useful information embedded in failed trajectories. To address this challenge, we propose Correction-Oriented Policy Optimization (CIPO), a simple and effective extension to RLVR that converts on-policy failed trajectories into correction-oriented supervision, without relying on any external signals. By jointly optimizing correction samples derived from the model's own failed attempts together with the standard RLVR objective, CIPO improves learning effectiveness while explicitly enhancing the model's ability to correct its own errors. Extensive experiments across 11 benchmarks spanning mathematical reasoning and code generation demonstrate that CIPO consistently and significantly outperforms strong baselines in both reasoning and correction performance. Moreover, CIPO yields stronger pass@K gains, indicating that it improves the model's intrinsic reasoning capacity rather than merely redistributing probability mass over existing correct answers.

We introduce \textsc{Cattle Trade, a multi-agent benchmark for evaluating large language models (LLMs) as agents in strategic reasoning under imperfect information, adversarial interaction, and resource constraints. The benchmark combines auctions, hidden-offer trade challenges (TCs), bargaining, bluffing, opponent modeling, and resource allocation within a single long-horizon game lasting 50--60 turns. Unlike prior agent benchmarks that test these abilities in isolation, \textsc{Cattle Trade} evaluates whether agents integrate them across a competitive, multi-agent economic game with conflicting incentives. The benchmark logs every bid, TC offer, counteroffer, and card selection, enabling behavioural analysis beyond final scores or win rates. We evaluate seven cost-efficient language models and three deterministic code agents across 242 games. Strategic coherence, in particular spending efficiency, resource discipline, and phase-adaptive bidding, is associated with rank more strongly than spending volume or any single subskill. Two heuristic code agents outperform most tested LLMs, and behavioural traces surface recurring LLM failure modes including overbidding, self-bidding, bankrupt TC initiation, and weak opponent-state adaptation. Evaluating agentic competence requires benchmarks that test the joint deployment of multiple capabilities in multi-agent environments with conflicting incentives, uncertainty, and economic dynamics.

The increased use of Large Language Models (LLMs) in geography raises substantial questions about the safety of integrating these tools across a wide range of processes and analyses, given our very limited understanding of their inner workings. In this extended abstract, we examine how LLMs process relative geographic space using activation patching, an emerging tool for mechanistic interpretability.

Evaluating object removal in images and videos remains challenging because the task is inherently one-to-many, yet existing metrics frequently disagree with human perception. Full-reference metrics reward copy-paste behaviors over genuine erasure; no-reference metrics suffer from systematic biases such as favoring blurry results; and global temporal metrics are insensitive to localized artifacts within edited regions. To address these limitations, we propose RC (Removal Coherence), a pair of perception-aligned metrics: RC-S, which measures spatial coherence via sliding-window feature comparison between masked and background regions, and RC-T, which measures temporal consistency via distribution tracking within shared restored regions across adjacent frames. To validate RC and support community benchmarking, we further introduce PROVE-Bench, a two-tier real-world benchmark comprising PROVE-M, an 80-video paired dataset with motion augmentation, and PROVE-H, a 100-video challenging subset without ground truth. Together, RC metrics and PROVE-Bench form the PROVE (Perceptual RemOVal cohErence) evaluation framework for visual media. Experiments across diverse image and video benchmarks demonstrate that RC achieves substantially stronger alignment with human judgments than existing evaluation protocols. The code for RC metrics and PROVE-Bench are publicly available at: https://github.com/xiaomi-research/prove/.

This work reformulates language generation as a stochastic optimal control problem, providing a unified theoretical perspective to analyze autoregressive and diffusion models and explain their limitations (Efficiency-Fidelity Paradox, Irreversibility Error Propagation, Optimization Tractability and Fidelity) in terms of combination of trajectory singularity, adjoint state vanishing, and gradient absence. To address these issues, we approximate the solution to the Hamilton-Jacobi-Bellman (HJB) equation, yielding an optimal policy that acts as a closed-loop controller. To bypass the intractability of directly solving the HJB PDE, we employ Flow Matching as the optimal trajectory solver within the rectified latent control space. This allows our Manta-LM with Global Integral Operator to approximate the global vector field, effectively realizing a model that simultaneously achieves high-fidelity text generation and efficient, low-cost parallel sampling. Empirically, our method achieves strong performance on language modeling and conditional generation tasks, while exhibiting improved stability, efficiency, and controllability.

Developing machine learning interatomic potentials (MLIPs) for complex materials systems remains challenging because it requires expertise in atomistic simulations, machine learning, and workflow design, as well as iterative active learning procedures. Existing automated pipelines typically assume a fixed sequence of stages or depend on domain experts, which limits their adaptability to heterogeneous materials systems where the optimal curriculum is not known in advance. To lower the barrier to developing MLIPs for non-experts, we propose Lang2MLIP, a multi-agent framework that takes natural-language input and formulates end-to-end MLIP development as a sequential decision-making problem solved by large language models (LLMs). At each step, a decision-making agent observes the current dataset, model, evaluation results, and execution log, and then automatically selects an appropriate action to improve the model. This removes the need for a predefined pipeline and enables the agent to self-correct by revisiting earlier subsystems when new failures arise. We evaluate this approach on a solid electrolyte interphase (SEI) system with multiple components and interfaces. These results suggest that LLM-based multi-agent systems are a promising direction for automating MLIP development and making it more accessible to non-experts.

Recent studies have reported $\textit{saturation effects}$ and $\textit{multiple descent behavior}$ in large dimensional kernel ridge regression (KRR). However, these findings are predominantly derived under restrictive settings, such as inner product kernels on sphere or strong eigenfunction assumptions like hypercontractivity. Whether such behaviors hold for other kernels remains an open question. In this paper, we establish a broad, new family of large dimensional kernels and derive the corresponding convergence rates of the generalization error. As a result, we recover key phenomena previously associated with inner product kernels on sphere, including: $i)$ the $\textit{minimax optimality}$ when the source condition $s\le 1$; $ii)$ the $\textit{saturation effect}$ when $s>1$; $iii)$ a $\textit{periodic plateau phenomenon}$ in the convergence rate and a $\textit {multiple-descent behavior}$ with respect to the sample size $n$.

Layer normalization (LN) is a fundamental component in modern deep learning, but its per-sample centering and scaling introduce non-negligible inference overhead. RMSNorm improves efficiency by removing the centering operation, yet this may discard benefits associated with centering. This paper propose a framework to determine whether an LN in an arbitrary DNN can be replaced by RMSNorm without changing the model function. The key idea is to fold LN's centering operation into upstream general linear layers by enforcing zero-mean outputs through the column-centered constraint (CCC) and column-based weight centering (CBWC). We extend the analysis to arbitrary DNNs, define such LNs as foldable LNs, and develop a graph-based detection algorithm. Our analysis shows that many LNs in widely used architectures are foldable, enabling exact inference-time conversion and end-to-end acceleration of 2% to 12% without changing model predictions. Experiments across multiple task families further show that, when exact equivalence is partially broken in practical training settings, our method remains competitive with vanilla LN while improving efficiency.

Activation functions are central to deep networks, influencing non-linearity, feature learning, convergence, and robustness. This paper proposes the Adaptive Arctangent Gated Activation (ArcGate) function, a flexible formulation that generates a broad spectrum of activation shapes via a three-stage non-linear transformation. Unlike conventional fixed-shape activations such as ReLU, GELU, or SiLU, ArcGate uses seven learnable parameters per layer, allowing the neural network to autonomously optimize its non-linearity to the specific requirements of the feature hierarchy and data distribution. We evaluate ArcGate using ResNet-50 and Vision Transformer (ViT-B/16) architectures on three widely used remote sensing benchmarks: PatternNet, UC Merced Land Use, and the 13-band EuroSAT MSI multispectral dataset. Experimental results show that ArcGate consistently outperforms standard baselines, achieving a peak overall accuracy of 99.67% on PatternNet. Most notably, ArcGate exhibits superior structural resilience in noisy environments, maintaining a 26.65% performance lead over ReLU under moderate Gaussian noise (standard deviation 0.1). Analysis of the learned parameters reveals a depth-dependent functional evolution, where the model increases gating strength in deeper layers to enhance signal propagation. These findings suggest that ArcGate is a robust and adaptive general node activation function for high-resolution earth observation tasks.

Holistic evaluation scores capture overall output quality but do not distinguish whether a model reproduced the structural form of a user's request from whether it preserved the user's specific intent. We propose a dimension-level intent fidelity evaluation framework, applied here through a structured prompt ablation study across 2,880 outputs spanning three languages, three task domains, and six LLMs, that separately measures structural recovery and intent fidelity for each semantic dimension. This framework reveals a systematic structural-fidelity split: among Chinese-language outputs with complete paired scores, 25.7% received perfect holistic alignment scores (GA=5) while exhibiting measurable dimensional intent deficits; among English-language outputs, this proportion rose to 58.6%. Human evaluation confirmed that these split-zone outputs represent genuine quality deficits and that dimensional fidelity scores track human judgements more reliably than holistic scores do. A public-private decomposition of 2,520 ablation cells characterises when models successfully compensate for missing intent and when they fail, while proxy annotation distinguishes prior inferability from default recoverability. A weight-perturbation experiment shows that moderate misalignment is typically absorbed, whereas severe dimensional inversion is consistently harmful. These findings demonstrate that dimension-level intent fidelity evaluation is a necessary complement to holistic assessment when evaluating LLM outputs for user-specific tasks.

Diffusion-based video generation has advanced substantially in visual fidelity and temporal coherence, but practical deployment remains limited by the quadratic complexity of full attention. Training-free sparse attention is attractive because it accelerates pretrained models without retraining, yet existing online top-$p$ sparse attention still spends non-negligible cost on mask prediction and applies shared thresholds despite strong head-level heterogeneity. We show that these two overlooked factors limit the practical speed-quality trade-off of training-free sparse attention in Video DiTs. To address them, we introduce a head-wise adaptive framework with two plug-in components: Temporal Mask Reuse, which skips unnecessary mask prediction based on query-key drift, and Error-guided Budgeted Calibration, which assigns per-head top-$p$ thresholds by minimizing measured model-output error under a global sparsity budget. On Wan2.1-1.3B and Wan2.1-14B, our method consistently improves XAttention and SVG2, achieving up to 1.93 times speedup at 720P while maintaining competitive video quality and similarity metrics.

Generative Recommendation (GenRec) models reformulate recommendation as a sequence generation task, representing items as discrete Semantic IDs used symmetrically as both inputs and prediction targets. We identify a critical dual-stage information bottleneck in this design: (1) the Input Bottleneck, where lossy quantization degrades fine-grained semantics, while popularity bias skews the learned representations toward frequent items, and (2) the Output Bottleneck, where imprecise discrete targets limit supervision quality. To address these issues, we propose AsymRec, an asymmetric continuous-discrete framework that decouples input and output representations. Specifically, Multi-expert Semantic Projection (MSP) maps continuous embeddings into the Transformer's hidden space via expert-specialized projections, preserving semantic richness and improving generalization to infrequent items. Multi-faceted Hierarchical Quantization (MHQ) constructs high-capacity, structured discrete targets through multi-view and multi-level quantization with semantic regularization, preventing dimensional collapse while retaining fine-grained distinctions. Extensive experiments demonstrate that AsymRec consistently outperforms state-of-the-art generative recommenders by an average of 15.8 %. The code will be released.

Long-horizon household tasks demand robust high-level planning and sustained reasoning capabilities, which are largely overlooked by existing embodied AI benchmarks that emphasize short-horizon navigation or manipulation and rely on fixed task categories. We introduce LongAct, a benchmark designed to evaluate planning-level autonomy in long-horizon household tasks specified through free-form instructions. By abstracting away embodiment-specific low-level control, LongAct isolates high-level cognitive capabilities such as instruction understanding, dependency management, memory maintenance, and adaptive planning. We further propose HoloMind, a VLM-driven agent with a DAG-based long-horizon hierarchical planner, a Multimodal Spatial Memory for persistent world modeling, an Episodic Memory for experience reuse, and a global Critic for reflective supervision. Experiments with GPT-5 and Qwen3-VL models show that HoloMind substantially improves long-horizon performance while reducing reliance on model scale. Even top models achieve only 59% goal completion and 16% full-task success, underscoring the difficulty of LongAct and the need for stronger long-horizon planning in embodied agents.

While Retrieval-Augmented Generation (RAG) is increasingly adopted to ground Large Language Models (LLMs) in software artifacts, the optimal configuration of its components remains an open question for software engineering (SE) tasks. The lack of systematic guidance forces practitioners into costly, ad-hoc experimentation. This paper presents a comprehensive, component-wise empirical study that dissects the RAG pipeline, evaluating over 21 distinct models and methods. Our study systematically isolates and evaluates 4 query processing techniques, 7 retrieval models spanning sparse, dense, and hybrid paradigms, 4 context refinement methods, and 6 distinct generators. We test these components on a suite of 3 core SE tasks: code generation, summarization, and repair. Our empirical findings reveal a crucial insight: the retriever-side components, particularly the choice of the retrieval algorithm, often exert a more significant influence on final system performance than the selection of the generator model. Strikingly, the classic lexical retriever BM25 demonstrates exceptionally robust performance across diverse tasks. Our analysis provides a practical, data-driven roadmap for researchers and practitioners, offering clear guidance on prioritizing optimization efforts when constructing effective RAG systems for software engineering contexts.

Power electronics systems are increasingly exposed to cyber threats due to their integration with digital controllers and communication networks. However, an attacker-oriented metric is still lacking to quantify the extent to which a node can be pushed toward instability within a privilege-constrained action space. This letter proposes an impedance-based Attack Reachable Domain (ARD) framework that maps feasible adversarial actions to critical-eigenvalue migration through impedance reshaping. Based on the ARD, an Attack Penetration Index is defined to quantify node-level cyber-vulnerability by jointly characterizing the penetration of the nominal stability margin and the accessibility of successful destabilizing attacks within a privilege-constrained action space. To make the proposed assessment computable when inverter models are unavailable, a practical gray-box workflow is further established by integrating existing impedance identification and differentiable surrogate tools. Case studies on a 4-bus system and a modified IEEE 39-bus system show that coordinated cross-layer manipulations are markedly more damaging than isolated single-layer attacks, and that the proposed metric reveals vulnerability patterns that cannot be inferred from grid-strength indicators.

This paper proposes a fully dynamic Deep Reinforcement Learning (DRL) method for rebalancing dockless bike-sharing systems, overcoming the limitations of periodic, system-wide interventions. We model the service through a graph-based simulator and cast rebalancing as a Markov decision process. A DRL agent routes a single truck in real time, executing localized pick-up, drop-off, and charging actions guided by spatiotemporal criticality scores. Experiments on real-world data show significant reductions in availability failures with a minimal fleet size, while limiting spatial inequality and mobility deserts. Our approach demonstrates the value of learning-based rebalancing for efficient and reliable shared micromobility.

Large Language Model (LLM) agents increasingly serve as personal assistants and workplace collaborators, where their utility depends on memory systems that extract, retrieve, and apply information across long-running conversations. However, both existing memory systems and benchmarks are built around the dyadic, single-user setup, even though real deployments routinely span groups and channels with multiple users interacting with the agent and with each other. This mismatch leaves three properties of group memory unmeasured: (i) group dynamics that go beyond concatenated one-on-one chats, (ii) speaker-grounded belief tracking, where the per-user memory modeling is needed, and (iii) audience-adapted language, where Theory-of-Mind shifts produce role-specific vocabulary. We introduce GroupMemBench, a benchmark that exposes all three. A graph-grounded synthesis pipeline produces multi-party conversations with controllable reply structure and conditions each message on per-user personas and target audiences. An adversarial query pipeline then binds every question to a specific asker across six categories, spanning multi-hop reasoning, knowledge update, term ambiguity, user-implicit reasoning, temporal reasoning, and abstention, and iteratively searches challenging, realistic queries that reflect comprehensive memory capability. Benchmarking leading memory systems exposes a sharp collapse: the strongest one reaches only 46.0% average accuracy, with knowledge update at 27.1% and term ambiguity at 37.7%, while a simple BM25 baseline matches or exceeds most agent memory systems. This indicates current memory ingestion erases the structural and lexical features group memory depends on, leaving multi-user memory far from solved.

Offline-to-online reinforcement learning harnesses the stability of offline pretraining and the flexibility of online fine-tuning. A key challenge lies in the non-stationary distribution shift between offline datasets and the evolving online policy. Common approaches often rely on static mixing ratios or heuristic-based replay strategies, which lack adaptability to different environments and varying training dynamics, resulting in suboptimal tradeoff between stability and asymptotic performance. In this work, we propose Reinforcement Learning with Optimized Adaptive Data-mixing (ROAD), a dynamic plug-and-play framework that automates the data replay process. We identify a fundamental objective misalignment in existing approaches. To tackle this, we formulate the data selection problem as a bi-level optimization process, interpreting the data mixing strategy as a meta-decision governing the policy performance (outer-level) during online fine-tuning, while the conventional Q-learning updates operate at the inner level. To make it tractable, we propose a practical algorithm using a multi-armed bandit mechanism. This is guided by a surrogate objective approximating the bi-level gradient, which simultaneously maintains offline priors and prevents value overestimation. Our empirical results demonstrate that this approach consistently outperforms existing data replay methods across various datasets, eliminating the need for manual, context-specific adjustments while achieving superior stability and asymptotic performance.

Multimedia verification requires not only accurate conclusions but also transparent and contestable reasoning. We propose a contestable multi-agent framework that integrates multimodal large language models, external verification tools, and arena-based quantitative bipolar argumentation (A-QBAF) as a submission to the ICMR 2026 Grand Challenge on Multimedia Verification. Our method decomposes each case into claim-centered sections, retrieves targeted evidence, and converts evidence into structured support and attack arguments with provenance and strength scores. These arguments are resolved through small local argument graphs with selective clash resolution and uncertainty-aware escalation. The resulting system generates section-wise verification reports that are transparent, editable, and computationally practical for real-world multimedia verification. Our implementation is public at: https://github.com/Analytics-Everywhere-Lab/MV2026_the_liems.

Two-Stage Robust Optimization (2RO) with discrete uncertainty is challenging, often rendering exact solutions prohibitive. Scenario reduction alleviates this issue by selecting a small, representative subset of scenarios to enable tractable computation. However, existing methods are largely problem-agnostic, operating solely on the uncertainty set without consulting the feasible region or recourse structure. In this paper, we introduce PRISE, a problem-driven sequential lookahead heuristic that constructs reduced scenario sets by evaluating the marginal impact of each scenario. While PRISE yields high-quality scenario subsets, each selection step requires solving multiple subproblems, making it computationally expensive at scale. To address this, we propose NeurPRISE, a neural surrogate model built on a GNN-Transformer backbone that encodes the per-scenario structure via graph convolution and captures cross-scenario interactions through attention. NeurPRISE is trained via imitation learning with a gain-aware ranking objective, which distills marginal gain information from PRISE into a learned scoring function for scenario ranking and selection. Extensive results on three 2RO problems show that NeurPRISE consistently achieves competitive regret relative to comprehensive methods, maintains strong calability with varying numbers of scenarios, and delivers 7-200x speedup over PRISE. NeurPRISE also exhibits strong zero-shot generalization, effectively handling instances with larger problem scales (up to 5x), more scenarios (up to 4x), and distribution shifts.

Building black-box models for dynamical systems from data is a challenging problem in machine learning, especially when asymptotic stability guarantees are required. In this paper, we introduce a novel stability-ensuring and backpropagation-compatible projection scheme based on the Schur decomposition for the state matrix of linear discrete-time state-space layers, as well as an alternative pre-factorized formulation of the methodology. The proposed methods dynamically project the quasi-triangular factor of the state matrix's real Schur decomposition onto its nearest stable peer, ensuring stable dynamics with minimal overparameterization. Experiments on synthetic linear systems demonstrate that the method achieves accuracy and convergence rates comparable to those of state-of-the-art stable-system identification techniques, despite a marginal increase in computational complexity. Furthermore, the lower weight count facilitates convergence during training without sacrificing accuracy in stacked neural-network architectures with static nonlinearities targeting real-world datasets. These results suggest that the Schur-based projection provides a numerically robust framework for identifying complex dynamics on par with the State of the Art while satisfying strict asymptotic-stability requirements.

Large Language Models (LLMs) augmented with Retrieval-Augmented Generation (RAG) techniques are revolutionizing applications across multiple domains, such as healthcare, finance, and customer service. Despite their potential, evaluating RAG systems remains a complex challenge due to the stochastic nature of generated outputs and the intricate interplay between retrieval and generation components. This paper introduces Deepchecks, a comprehensive framework tailored for evaluating RAG applications. Deepchecks' evaluation framework addresses RAG applications evaluation through a multi-faceted approach, root cause analysis and production monitoring. By ensuring alignment with application-specific requirements, Deepchecks framework provides a robust foundation for assessing reliability, relevance, and user satisfaction in RAG systems.

Autoregressive video diffusion models support real-time synthesis but suffer from error accumulation and context loss over long horizons. We discover that attention heads in AR video diffusion transformers serve functionally distinct roles as local heads for detail refinement, anchor heads for structural stabilization, and memory heads for long-range context aggregation, yet existing methods treat them uniformly, leading to suboptimal KV cache allocation. We propose Head Forcing, a training-free framework that assigns each head type a tailored KV cache strategy: local and anchor heads retain only essential tokens, while memory heads employ a hierarchical memory system with dynamic episodic updates for long-range consistency. A head-wise RoPE re-encoding scheme further ensures positional encodings remain within the pretrained range. Without additional training, Head Forcing extends generation from 5 seconds to minute-level duration, supports multi-prompt interactive synthesis, and consistently outperforms existing baselines. Project Page: https://jiahaotian-sjtu.github.io/headforcing.github.io/.

Large language models (LLMs) have become a strong foundation for multi-agent systems, but their effectiveness depends heavily on orchestration design. Across different tasks, role design, capacity assignment, and dependency construction jointly affect both solution quality and execution efficiency. Existing approaches automate parts of this design process, yet they often optimize these decisions partially or sequentially, and rely on execution-level feedback that provides limited credit assignment for local orchestration decisions. We propose LEMON (\textbf{L}earning \textbf{E}xecutable \textbf{M}ulti-agent \textbf{O}rchestratio\textbf{N} via Counterfactual Reinforcement Learning), an LLM-based orchestrator that generates an executable orchestration specification. The specification integrates task-specific roles, customized duties, capacity levels, and dependency structure into a single deployable system. To train the orchestrator, we augment the orchestration-level GRPO objective with a localized counterfactual signal that edits role, capacity, or dependency fields and applies the resulting reward contrast only to the edited spans. Experiments on six reasoning and coding benchmarks, including MMLU, GSM8K, AQuA, MultiArith, SVAMP, and HumanEval, show that LEMON achieves state-of-the-art performance among the evaluated multi-agent orchestration methods. Our code is available at https://anonymous.4open.science/r/LEMON-B23C.

Purpose: This study investigates the transcription principles underlying Huìtóngguǎnxì Huáyíyìyǔ (HHY), a series of multilingual glossaries compiled by the Ming government between the fifteenth and sixteenth centuries for interpreter training. The study treats HHY not as a collection of isolated language materials, but as a coherent multilingual transcription system representing spoken forms of non-Chinese languages through Chinese characters. Methods: A substantial portion of HHY was digitized and aligned with Chinese phonological categories. Previous reconstructions of individual language sections were critically reviewed and integrated into a unified comparative database. The analysis focuses on cross-linguistic regularities in Main Transcription (MT) and Supplementary Transcription (ST) across eight language sections. Results: MT generally represents sounds compatible with the Chinese syllable structure of the period, whereas ST mainly encodes phonetic features less compatible with Chinese phonology. The analysis further shows that Chinese phonological categories were used more flexibly in foreign-language transcription than previously assumed. HHY therefore functioned as a relatively systematic method of phonetic approximation rather than a direct projection of Chinese phonology onto non-Chinese languages. Conclusion: HHY can be analyzed as an internally structured transcription system rather than merely as a collection of glossaries. More broadly, the study demonstrates that historical transcription systems can provide valuable evidence for historical phonology, particularly for under-documented Asian languages with limited historical records.

Context: Retrieval-augmented code generation relies on cross-file repository context, but retrieved snippets may come from obsolete project states. Objectives: We study whether temporally stale repository snippets act as harmless noise or actively induce current-state-incompatible code. Methods: We conduct a controlled diagnostic study on a curated 17-sample set of production-helper signature changes from five Python repositories. For each sample, we compare current-only, stale-only, no-retrieval, and mixed current/stale retrieval conditions under prompts that hide commit freshness and expected current signatures. Results: Under neutralized prompts, stale-only retrieval induces stale helper references on 15/17 Qwen2.5-Coder-7B-Instruct samples and 13/17 gpt-4.1-mini samples, corresponding to 88.2 and 76.5 percentage-point increases over current-only retrieval. No retrieval produces zero stale references but only 1/17 passing completions. The two models share 75.0% Jaccard overlap among stale-triggering samples, and mixed conditions show that adding valid current evidence largely rescues stale-only failures. Conclusion: Temporal validity of retrieved repository context is a distinct diagnostic variable for Code RAG robustness: stale context can actively bias models toward obsolete repository state rather than merely removing useful evidence.

We introduce EvoLib, a test-time learning framework that enables large language models to accumulate, reuse, and evolve knowledge across problem instances without parameter updates or external supervision. Instead of adapting model parameters, our approach maintains a shared library of knowledge abstractions, including modular skills and reflective insights, automatically extracted from the model's own inference trajectories. To support continual improvement, we introduce a principled weighting and consolidation mechanism that jointly optimizes for immediate utility and long-term value. This allows simple, instance-specific abstractions to evolve into more general and reusable ones over time. Across challenging benchmarks in mathematical reasoning, code generation, and multi-turn agentic environments, EvoLib improves substantially over the top test-time scaling and learning methods without ground-truth feedback.

The Context-Compliance Regime in Retrieval-Augmented Generation (RAG) occurs when retrieved context dominates the final answer even when it conflicts with the model's parametric knowledge. Accuracy alone does not reveal how retrieved context causally shapes answers under such conflict. We introduce Context-Driven Decomposition (CDD), a belief-decomposition probe that operates at inference time and serves as an intervention mechanism for controlled retrieval conflict. Across Epi-Scale stress tests, TruthfulQA misconception injection, and cross- model reruns, CDD exposes three patterns. P1: context compliance is measurable in an upper-bound adversarial setting, where Standard RAG reaches 15.0% accuracy on TruthfulQA misconception injection (N=500). P2: adversarial accuracy gains transfer across model families: CDD improves accuracy on Gemini-2.5-Flash and on Claude Haiku/Sonnet/Opus, but rationale-answer causal coupling does not transfer. CDD reaches 64.1% mistake- injection causal sensitivity on Gemini-2.5-Flash, while sensitivities for all three Claude variants fall in the [-3%, +7%] range, suggesting that the Claude-side accuracy gains operate through a mechanism distinct from the explicit conflict-resolution trace. P3: explicit conflict decomposition improves robustness under temporal drift and noisy distractors, with CDD reaching 71.3% on temporal shifts and 69.9% on distractor evidence on the full Epi-Scale adversarial benchmark. These three patterns identify context-compliance as a structural axis along which standard RAG can be probed and intervened on, distinct from retrieval-quality or single-method robustness questions, and motivate releasing Epi-Scale for systematic study across model families and retrieval pipelines.

We propose a new method of learning from positive and unlabeled (PU) examples in highly imbalanced datasets. Many real-world problems, such as disease gene identification, targeted marketing, fraud detection, and recommender systems, are hard to address with machine learning methods, due to limited labeled data. Often, training data comprises positive and unlabeled instances, the latter typically being dominated by negative, but including also several positive instances. While PU learning is well-studied, few methods address imbalanced settings or hard-to-detect positive examples that resemble negative ones. Our approach uses a focused empirical risk estimator, incorporating both positive and unlabeled examples to train binary classifiers. Empirical evaluations demonstrate state-of-the-art performance on imbalanced datasets under two labeling mechanisms - selecting positives completely at random (SCAR) and selecting at random (SAR). Beyond these controlled experiments, we demonstrate the value of the proposed method in the real-world application of financial misstatement detection.

Multi-step LLM reasoning over structured tables fails because planning and execution share no explicit cell-grounding contract. Existing methods constrain the planner to a left-to-right factorization at odds with table permutation invariance, and score intermediate states by generated content alone, overlooking cell grounding. We conduct a pilot study showing that diffusion language models (DLMs) produce more human-aligned and permutation-stable cell attention on tables than autoregressive models, with a 40.2% median reduction in attention-AUROC variability under row reordering. Motivated by this, we propose TABALIGN, a planned table reasoning framework that operationalizes the contract. TABALIGN pairs a masked DLM planner, whose bidirectional denoising emits plan steps as binary cell masks, with TABATTN, a lightweight verifier trained on 1,600 human-verified attention standards to score each step by its attention overlap with the plan-designated mask. Across eight benchmarks covering table question answering and fact verification, TABALIGN improves average accuracy by 15.76 percentage points over the strongest open-source baseline at comparable 8B-class scale, with a matched-backbone ablation attributing 2.87 percentage points of this gain to the DLM planner over an AR planner on a fixed reasoner. Cleaner DLM plans also accelerate downstream reasoning execution by 44.64%.

Autonomous agents powered by Large Language Models (LLMs) acquire external functionalities through third-party skills available in open marketplaces. Adopting these integrations broadens the potential attack surface, prompting a need for systematic security evaluation. Current auditing mechanisms are effective at identifying explicit code payloads and predefined threat contents through security scanning. These detection mechanisms are bypassed if malicious behaviors lack direct injection and are instead synthesized dynamically at runtime through the agent's inherent generative capabilities. Exploring this blind spot, we introduce Semantic Compliance Hijacking (SCH), a payload-less supply chain attack targeting autonomous coding environments. The SCH approach translates malicious goals into unstructured natural language instructions formatted as necessary compliance rules, leading the agent to generate and execute unauthorized code. To assess the real-world viability of this attack, we developed an automated pipeline to evaluate its effectiveness across a test matrix comprising three mainstream agent frameworks and three distinct foundation models using contextualized scenarios. The findings demonstrate the pervasive nature of this threat, with SCH achieving peak success rates of up to 77.67% for confidentiality breaches and 67.33% for Remote Code Execution (RCE) under the most vulnerable configurations. Furthermore, the introduction of Multi-Skill Automated Optimization (MS-AO) further boosted attack efficacy. By omitting recognizable Abstract Syntax Tree (AST) signatures and explicit harmful intents, the manipulated skill files maintained a 0.00% detection rate, evading current scanning tools. This research highlights an underexplored attack surface within agent supply chains, pointing to a necessary transition from signature-based detection models toward semantic intent validation.

Omni-modal large language models have demonstrated remarkable potential in holistic multimodal understanding; however, the token explosion caused by high-resolution audio and video inputs remains a critical bottleneck for real-time applications and long-form reasoning. Existing omni-modal token compression methods typically prune tokens at the input embedding level, relying on audio-video similarity or temporal co-occurrence as proxies for semantic relevance. In practice, such assumptions are often unreliable. To address this limitation, we propose OmniDrop, a training-free, layer-wise token pruning framework that progressively prunes audiovisual tokens within the LLM decoder layers rather than at the input-level, allowing early layers to preserve sufficient omni-modal information fusion before aggressively removing tokens in deeper layers. We further utilize text queries as guidance for modality-agnostic and task-adaptive token pruning. We also introduce a temporal diversity score that encourages balanced token survival to preserve global temporal context. Experimental results across various audiovisual benchmarks demonstrate that OmniDrop outperforms all baselines by up to 3.58 points while reducing prefill latency by up to 40% and memory usage by up to 14.7%.

Chain-of-Thought (CoT) reasoning has become a foundation for eliciting multi-step reasoning in large language models, but recent studies show that its benefits do not scale monotonically with chain length: while longer CoT generally enables a model to tackle harder problems, on a given problem, accuracy typically increases with CoT length up to a point, after which it declines. We identify a major cause of this phenomenon: as the CoT grows, the model's attention to critical insights produced earlier in the trace gradually weakens, making those insights progressively less accessible when they are most needed. Therefore, we propose \textbf{InsightReplay}, a stateful reasoning approach in which the model periodically extracts critical insights from its reasoning trace and replays them near the active generation frontier, keeping them accessible as the reasoning scales. Extensive experiments on a $\mathbf{2}\!\times\!\mathbf{3}\!\times\!\mathbf{4}$ benchmark grid, covering model scales $\{\text{8B}, \text{30B}\}$, model families $\{\text{Qwen3.5}, \text{DeepSeek-R1-Distill-Qwen}, \text{Gemma-4}\}$, and reasoning benchmarks $\{\text{AIME}, \text{HMMT}, \text{GPQA Diamond}, \text{LiveCodeBench v5}\}$, show that 3-round InsightReplay yields accuracy gains across \textbf{all 24 settings}, with an averaged improvement of $\mathbf{+1.65}$ points over standard CoT, and a largest single-setting gain of $\mathbf{+9.2}$ points on R1-Distill-32B's LiveCodeBench v5 subset. Our results suggest that the effectiveness of test-time scaling depends not only on how much a model reasons, but also on whether critical intermediate insights remain accessible throughout long reasoning trajectories.

The Intelligence Impact Quotient (IIQ) is a composite metric intended to quantify the depth to which AI systems are integrated into organizational work and their impact. Rather than treating access counts or aggregate token volume as sufficient evidence of impact, IIQ combines a novelty-weighted, time-decayed token stock with usage frequency, a grace-period recency gate, organizational leverage, task complexity, and autonomy. The formulation produces a raw Intelligence Adoption Index (IAI) and a normalized 0-1000 IIQ index for comparison between heterogeneous users and units. We also derive sub-daily update rules and a bounded interpretation layer for estimated efficiency and financial impact. The paper positions IIQ as a deployment-oriented measurement framework: a formal proposal for tracking AI embedding in workflows, not a direct measure of model capability or a substitute for causal productivity evaluation. Synthetic scenarios illustrate how the revised metric distinguishes between frequent low-leverage use, semantically repetitive prompting, and more autonomous, higher-consequence AI-assisted work.

As AI agents move from chat interfaces to systems that read private data, call tools, and execute multi-step workflows, guardrails become a last line of defense against concrete deployment harms. In these settings, guardrail failures are no longer merely answer-quality errors: they can leak secrets, authorize unsafe actions, or block legitimate work. The hardest failures are often contextual: whether an action is acceptable depends on local privacy norms, organizational policies, and user expectations that resist pre-deployment specification. This creates a practical gap: guardrails must adapt to their own operating environments, yet deployment feedback is typically limited to sparse, noisy user-reported failures, and repeated fine-tuning is often impractical. To address this gap, we propose LiSA (Lifelong Safety Adaptation), a conservative policy induction framework that improves a fixed base guardrail through structured memory. LiSA converts occasional failures into reusable policy abstractions so that sparse reports can generalize beyond individual cases, adds conflict-aware local rules to prevent overgeneralization in mixed-label contexts, and applies evidence-aware confidence gating via a posterior lower bound, so that memory reuse scales with accumulated evidence rather than empirical accuracy alone. Across PrivacyLens+, ConFaide+, and AgentHarm, LiSA consistently outperforms strong memory-based baselines under sparse feedback, remains robust under noisy user feedback even at 20% label-flip rates, and pushes the latency--performance frontier beyond backbone model scaling. Ultimately, LiSA offers a practical path to secure AI agents against the unpredictable long tail of real-world edge risks.

Hallucination detection in large language models (LLMs) requires balancing accu racy, efficiency, and robustness to distribution shift. Black-box consistency methods are effective but demand repeated inference; single-pass white-box probes are effi cient yet treat answer representations in isolation, often degrading sharply under domain shift. We propose QAOD (Question-Answer Orthogonal Decomposition), a single-pass framework that projects away the question-aligned direction from the answer representation to obtain a question-orthogonal component that suppresses domain-conditioned variation. To identify informative signals, QAOD further selects layers via diversity-penalized Fisher scoring and discriminative neurons via Fisher importance. To address both in-domain detection and cross-domain generalization, we design two complementary probing strategies: pairing the or thogonal component with question context yields a joint probe that maximizes in-domain discriminability, while using the orthogonal component alone preserves domain-agnostic factuality signals for robust transfer. QAOD's joint probe achieves the best in-domain AUROC across all evaluated model-dataset pairs, while the orthogonal-only probe delivers the strongest OOD transfer, surpassing the best white-box baseline by up to 21% on BioASQ at under 25% of generation cost.

Multimodal large language models (MLLMs) have emerged as a powerful backbone for multimodal embeddings. Recent methods introduce chain-of-thought (CoT) reasoning into the embedding pipeline to improve retrieval quality, but remain costly in both model size and inference cost. They typically employ separate reasoner and embedder with substantial parameter overhead, and generate CoT indiscriminately for every input. However, we observe that for simple inputs, discriminative embeddings already perform well, and redundant reasoning can even mislead the model, degrading performance. To address these limitations, we propose Think When Needed (TWN), a unified multimodal embedding framework with adaptive reasoning. TWN introduces a dual-LoRA architecture that attaches reasoning and embedding adapters to a shared frozen backbone, detaching gradients at their interface to mitigate gradient conflicts introduced by joint optimization while keeping parameters close to a single model. Building on this, an adaptive think mechanism uses a self-supervised routing gate to decide per input whether to generate CoT, skipping unnecessary reasoning to reduce inference overhead and even improve retrieval quality. We further explore embedding-guided RL to optimize CoT quality beyond supervised training. On the 78 tasks of MMEB-V2, TWN achieves state-of-the-art embedding quality while being substantially more efficient than existing generative methods, requiring only 3-5% additional parameters relative to the backbone and up to 50% fewer reasoning tokens compared to the full generative mode.

Many real-world coding challenges are open-ended and admit no known optimal solution. Yet, recent progress in LLM coding has focused on well-defined tasks such as feature implementation, bug fixing, and competitive programming. Open-ended coding remains a weak spot for LLMs, largely because open-ended training problems are scarce and expensive to construct. Our goal is to synthesize open-ended coding problems at scale to train stronger LLM coders. We introduce FrontierSmith, an automated system for iteratively evolving open-ended problems from existing closed-ended coding tasks. Starting from competitive programming problems, FrontierSmith generates candidate open-ended variants by changing the problems'goals, restricting outputs, and generalizing inputs. It then uses a quantitative idea divergence metric to select problems that elicit genuinely diverse approaches from different solvers. Agents then generate test cases and verifiers for the surviving candidates. On two open-ended coding benchmarks, training on our synthesized data yields substantial gains over the base models: Qwen3.5-9B improves by +8.82 score on FrontierCS and +306.36 (Elo-rating-based performance) on ALE-bench; Qwen3.5-27B improves by +12.12 and +309.12, respectively. The synthesized problems also make agents take more turns and use more tokens, similar to human-curated ones, suggesting that closed-ended seeds can be a practical starting point for long-horizon coding data.

The shift toward interacting with frozen, "black-box" Large Language Models (LLMs) has transformed prompt engineering from a heuristic exercise into a critical optimization challenge. We propose a Reinforcement Learning (RL) framework for training learned prompting policies via iterative distillation of experience. In this architecture, a lightweight prompter model is optimized to maximize task-specific rewards for a larger, frozen worker LLM. By utilizing a contrastive experience buffer that couples scalar rewards with dense textual critiques, our approach effectively amortizes iterative prompt refinement into single-shot policy weights. Our experimental analysis focuses on the Big Bench Extra Hard (BBEH) and Tau-bench suites, covering a diverse range of multi-step reasoning and tool-use tasks. We demonstrate significant gains, improving performance from 55% to 90% in logic-intensive reasoning and 74% to 91% in tool-use tasks. Furthermore, we analyze the structural evolution of prompts, demonstrating how the policy discovers specialized algorithmic heuristics. We provide comprehensive comparisons against state-of-the-art evolutionary baselines like GEPA, showing that iterative distillation achieves superior performance with higher sample efficiency.

Partially Observable Markov Decision Processes (POMDPs) are the standard framework for decision-making under uncertainty. While sampling-based methods scale well, they lack formal correctness guarantees, making them unsuitable for safety-critical applications. Conversely, formal synthesis techniques provide correctness-by-construction but often struggle with scalability, as general POMDP synthesis is undecidable. To bridge this gap, we propose a synthesis framework that integrates sampling, automata learning, and model-checking. Inspired by Angluin's $L^*$ algorithm, our approach utilizes sampling as a membership oracle and model-checking as an equivalence oracle. This enables the synthesis of finite-state controllers with formal guarantees, provided the sampling-induced policy is regular. We establish a relative completeness result for this framework. Experimental results from our prototypical implementation demonstrate that this method successfully solves threshold-safety problems that remain challenging for existing formal synthesis tools. We believe our algorithm serves as a valuable component in a portfolio approach to tackling the inherent difficulty of POMDP synthesis problems.

Mixture-of-Experts (MoE) architectures enhance the efficiency of large language models by activating only a subset of experts per token. However, standard MoE employs a fixed Top-K routing strategy, leading to redundant computation and suboptimal inference latency. Existing acceleration methods either require costly retraining with architectural changes or suffer from severe performance drop at high sparsity due to train-inference mismatch. To address these limitations, we propose BEAM (Binary Expert Activation Masking), a novel method that learns token-adaptive expert selection via trainable binary masks. With a straight-through estimator and an auxiliary regularization loss, BEAM induces dynamic expert sparsity through end-to-end training while maintaining model capability. We further implement an efficient custom CUDA kernel for BEAM, ensuring seamless integration with the vLLM inference framework. Experiments show that BEAM retains over 98\% of the original model's performance while reducing MoE layer FLOPs by up to 85\%, achieving up to 2.5$\times$ faster decoding and 1.4$\times$ higher throughput, demonstrating its effectiveness as a practical, plug-and-play solution for efficient MoE inference.

Generative retrieval offers a promising alternative by unifying the fragmented multi-stage retrieval process into a single end-to-end model. However, its practical adoption in industrial e-commerce search remains challenging, given the massive and dynamic product catalogs, strict latency requirements, and the need to align retrieval with downstream ranking goals. In this work, we propose a retrieval framework tailored for real-world recall scenarios, positioning generative retrieval as a recall-stage supplement rather than an end-to-end replacement. Our method, CQ-SID (Category-and-Query constrained Semantic ID), employs category-aware and query-item contrastive learning along with Residual Quantized VAEs to encode items into hierarchical semantic cluster identifiers, significantly reducing beam search complexity. Additionally, we develop EG-GRPO (Expert-Guided Group Relative Policy Optimization), a reinforcement learning approach that aligns generative recall with downstream ranking under sparse rewards by injecting ground-truth samples to stabilize training. Offline experiments on TmallAPP search logs show that CQ-SID achieves up to 26.76% and 11.11% relative gains in semantic and personalized click hitrate over RQ-VAE baselines, while halving beam search size. EG-GRPO further improves multi-objective performance. Online A/B tests confirm gains in GMV (+1.15%) and UCTCVR (+0.40%). The generative recall channel now contributes substantially in production, accounting for over 50.25% of exposures, 58.96% of clicks, and 72.63% of purchases, demonstrating a viable path for deploying generative retrieval in real-world e-commerce systems.

Library fuzzing is essential for hardening the software supply chain, but adopting it at scale remains expensive. Practitioners still spend substantial effort on environment setup, struggle to generate harnesses that respect intricate API constraints, and lack reliable means to tell genuine library bugs from harness-induced crashes. Recent LLM-based systems automate parts of this pipeline, yet they typically operate as one-shot code generators that ignore runtime feedback, which limits both the depth of code they reach and the validity of the bugs they report. We argue that effective library fuzzing is iterative by nature: each campaign exposes new coverage bottlenecks and crashes, and the next campaign should evolve from these signals rather than restart from scratch. Building on this insight, we present FuzzAgent, a multi-agent system that turns library fuzzing into an evolutionary process, in which a team of specialized agents collaborates over the full fuzzing lifecycle and grounds every decision in concrete runtime evidence, so that the harness suite is successively refined toward deeper coverage and higher-fidelity crash analysis across rounds. We evaluate FuzzAgent on 20 real-world C/C++ libraries against four state-of-the-art baselines (OSS-Fuzz, OSS-Fuzz-Gen, PromptFuzz, and PromeFuzz). FuzzAgent completes the full fuzzing lifecycle for all 20 libraries without human intervention and reaches 179619 branches, exceeding OSS-Fuzz, PromptFuzz, PromeFuzz, and OSS-Fuzz-Gen by 45.1%, 73.2%, 92.1%, and 191.2%, respectively. FuzzAgent also identifies 102 genuine library bugs, 78 of which have already been acknowledged and fixed by upstream maintainers.

In hybrid automatic speech recognition (ASR) systems, the vocabulary size is unambiguous, typically determined by the number of phones, bi-phones, or tri-phones present in the language. In contrast, end-to-end ASR systems derive their vocabulary, often referred to as tokens from the text corpus used for training. The choice and, more importantly, the size of this vocabulary is a critical hyper-parameter in training end-to-end ASR systems. Tokenization algorithms such as Byte Pair Encoding (BPE), WordPiece, and Unigram Language Model (ULM) use the vocabulary size as an input hyper-parameter to generate the sub-words employed during ASR training. Popular toolkits like ESPNet provide a fixed vocabulary size in their training recipes, but there is little documentation or discussion in the literature regarding how these values are determined. Recent work [1] has formalized an approach to identify the vocabulary size best suited for end-to-end ASR, introducing a cost function framework that treats the tokenization process as a black box. In this paper, we build upon that foundation by curve fitting the training data and using the principle of first and second derivative tests in calculus to formally estimate the vocabulary size hyper-parameter. We demonstrate the utility and usefulness of our approach by applying it on a standard Librispeech corpus and show that the optimal choice of vocabulary size hyper-parameter improves the performance of the ASR. The main contribution of this paper in formalizing an approach to identify the vocabulary size best suited for training an end-to-end ASR system.

Reliable precipitation monitoring is essential for disaster risk reduction, water resources management, and agricultural decision-making. Multi-source satellite observations, particularly the combination of geostationary infrared and passive microwave measurements, have become a primary means of precipitation detection. Traditional multi-source satellite precipitation estimation methods remain computationally inefficient, and many deep learning methods lack the flexibility to incorporate new sensors without retraining the full model. Here we introduce PRISMA (Precipitation Inference from Satellite Modalities via generAtive modeling), a plug-and-play latent generative framework for multi-sensor precipitation estimation. PRISMA learns an unconditional precipitation prior from IMERG Final fields and constrains it through independently trained, sensor-specific conditional branches, allowing new observation sources to be incorporated without retraining the generative backbone. Applied to FY-4B AGRI infrared and GPM GMI microwave observations, PRISMA improves Critical Success Index by up to 40.3% and reduces root-mean-square error by 22.6% relative to infrared-only estimation within microwave swaths, while also improving probabilistic skill and maintaining an average inference time of about 37 s. Independent rain-gauge validation across China confirms consistent gains, and typhoon case studies show that microwave conditioning restores eyewall and spiral rainband structures, reducing storm-core mean absolute error by up to 42.3%. PRISMA thus provides an extensible and efficient framework for multi-sensor precipitation estimation.

Despite the popularity of the actor-critic method and the practical needs of collaborative policy training, existing works typically either overlook environmental heterogeneity or give up personalization altogether by training a single shared policy across all agents. We consider a federated actor-critic framework in which agents share a common linear subspace representation while maintaining personalized local policy components, and agents iteratively estimate the common subspace, local critic heads, and local policies (i.e., actors). Under canonical single-timescale updates with Markovian sampling, we establish finite-time convergence via a novel joint linear approximation framework. Specifically, we show that the critic error converges to zero at the rate of $\tilde{\mathcal{O}}(1/((1-γ)^4\sqrt{TK}))$, and the policy gradient norm converges to zero at the rate of $\tilde{\mathcal{O}}(1/((1-γ)^6\sqrt{TK}))$, where $T$ is the number of rounds, $K$ is the number of agents, and $γ\in (0,1)$ is the discount factor. These results demonstrate linear speedup with respect to the number of agents $K$, despite heterogeneous Markovian trajectories under distinct transition kernels and coupled learning dynamics. To address these challenges, we develop a new perturbation analysis for the projected subspace updates and QR decomposition steps, together with conditional mixing arguments for heterogeneous Markovian noise. Furthermore, to handle the additional complications induced by policy updates and temporal dependence, we establish fine-grained characterizations of the discrepancies between function evaluations under Markovian sampling and under temporally frozen policies. Experiments instantiate the framework within PPO on federated \texttt{Hopper-v5} action-map heterogeneity, showing gains over Single PPO and FedAvg PPO and downstream transfer from the learned shared trunk.

Time series forecasting has become increasingly critical in real-world scenarios, where future sequences are influenced not only by historical patterns but also by forthcoming events. In this context, forecasting must dynamically adapt to complex and stochastic future conditions, which introduces fundamental challenges in both forecasting and evaluation. Traditional methods typically rely on historical data or factual future conditions, while overlooking counterfactual scenarios. Furthermore, many existing approaches are restricted to simple structured conditions, limiting their ability to generalize to the real-world complexities. To address these gaps, we introduce the task of counterfactual time series forecasting with textual conditions, enabling more flexible and condition-aware forecasting. We propose a comprehensive evaluation framework that encompasses both factual and counterfactual settings, even in the absence of ground truth time series. Additionally, we present a novel text-attribution mechanism that distinguishes mutable from immutable factors, thereby improving forecast accuracy under sophisticated and stochastic textual conditions. The project page is at https://seqml.github.io/TADiff/

We introduce MemLineage, a defense for LLM agent memory that attaches both cryptographic provenance and LLM-mediated derivation lineage to every entry. Recent and concurrent work shows that untrusted content can be written into persistent agent state and re-enter later sessions as an instruction; the remaining systems question is how to preserve useful memory recall while preventing such state from justifying sensitive actions. MemLineage treats this as a chain-of-custody problem rather than a filtering problem. It is a six-module design around an RFC-6962 Merkle log over per-principal Ed25519-signed entries: a weighted derivation DAG records which retrieved entries influenced each new memory, and a max-of-strong-edges propagation rule makes Untrusted-Path Persistence hold for any chain whose attribution edges remain above threshold. The sensitive-action gate then refuses dispatches whose active justification descends from an external ancestor, while still allowing benign recall. We evaluate three defense cells against three memory-poisoning workloads on a deterministic mechanism-isolation harness; MemLineage is the only configuration in that harness that drives all three columns to zero ASR, while sub-millisecond per-operation overhead keeps it well below the noise floor of any LLM call. A Codex-backed AgentDojo bridge further separates strong-model behavior from defense-layer behavior: under an intentionally vulnerable tool-output profile, no-defense and signature-only baselines fail on all six banking pairs, while all MemLineage rows reduce strict AgentDojo ASR to zero. The core deterministic artifacts are byte-equal CI-verified; hosted-model AgentDojo and live-model sweeps are recorded as auditable logs rather than byte-pinned artifacts.

Current Large Language Models (LLMs) typically rely on coarse-grained national labels for pluralistic value alignment. However, such macro-level supervision often obscures intra-country value heterogeneity, yielding a loose alignment. We argue that resolving this limitation requires shifting from national labels to multi-dimensional demographic constraints, which can identify groups with predictable, high-consensus value preference. To this end, we propose DVMap (High-Consensus Demographic-Value Mapping), a framework for fine-grained pluralistic value alignment. In this framework, we first present a demographic archetype extraction strategy to construct a high-quality value alignment corpus of 56,152 samples from the World Values Survey (WVS) by strictly retaining respondents with consistent value preferences under identical demographics. Over this corpus, we introduce a Structured Chain-of-Thought (CoT) mechanism that explicitly guides LLMs to reason about demographic-value correlations. Subsequently, we employ Group Relative Policy Optimization (GRPO) to achieve adaptive anchoring of value distributions. To rigorously evaluate generalization, we further establish a triple-generalization benchmark (spanning cross-demographic, cross-country, and cross-value) comprising 21,553 samples. Experimental results demonstrate that DVMap effectively learns the manifold mapping from demographics to values, exhibiting strong generalization and robustness. On cross-demographic tests, Qwen3-8B-DVMap achieves 48.6% accuracy, surpassing the advanced open-source LLM DeepSeek-v3.2 (45.1%). The source code and dataset are available at https://github.com/EnlightenedAI/DVMap.

"Oh-Oh, yes, I'm the great pretender. Pretending that I'm doing well. My need is such, I pretend too much..." summarizes the state in the area of jailbreak creation and evaluation. You find this method to generate adversarial attacks proposed by a reputable institution (e.g., BoN from Anthropic or Crescendo from Microsoft Research). However, this method does not deliver on the promise claimed in the paper despite having top ASR scores against industry-grade LLMs. You successfully generate the jailbreak prompts against your target (open) model. However, the generated jailbreak prompt works against the target model with a 50% consecutive success rate (5 out of 10 attempts) despite having an 80% ASR (on paper) on the latest closed-source model (with a guardrail system)! This observation leads us to think. First, Attack Success Rate (ASR), the primary metric for LLM jailbreak benchmarking, is not a stable quantity. Second, published ASR numbers are therefore systematically inflated and incomparable across papers. Therefore, we wonder "Why a successful jailbreak prompt does not perform consistently well against a target model on which the prompts have been optimized?". To answer this question, we study the impact of stochasticity not only during attack evaluation but also during attack generation. Our evaluation includes several jailbreak attacks, models (different sizes and providers), and judges. In addition, we propose a new metric and two new frameworks (CAS-eval and CAS-gen). Our evaluation framework, CAS-eval, shows that an attack can have an ASR drop of up to 30 percentage points when a jailbreak prompt needs to succeed on more than one attempt. Thankfully, our attack generation framework (CAS-gen) improves previous jailbreak methods and helps them recover this loss of 30 percentage points!

The Capacitated Vehicle Routing Problem (CVRP) is a fundamental NP-hard problem with broad applications in logistics and transportation. Real-world CVRPs often involve diverse objectives and complex constraints, such as time windows or backhaul requirements, motivating the development of a unified solution framework. Recent reinforcement learning (RL) approaches have shown promise in combinatorial optimization, yet they rely on end-to-end learning and lack explicit problem-solving knowledge, limiting solution quality. In this paper, we propose a knowledge-embedded framework inspired by the Route-First Cluster-Second heuristics. It incorporates knowledge at two levels: (1) decomposing CVRPs into the route-first and cluster-second subproblems, and (2) leveraging dynamic programming to solve the second subproblem, whose results guide the RL-based constructive solver to solve the first problem. To mitigate partial observability caused by problem decomposition, we introduce a unified history-enhanced context processing module. Extensive experiments show that this framework achieves superior solution quality compared with state-of-the-art learning-based methods, with a smaller gap to classical heuristics, demonstrating strong generalization across diverse CVRP variants.

Coding agents powered by large language models are increasingly expected to perform realistic software maintenance tasks beyond isolated issue resolution. Existing benchmarks have shifted toward realistic software evolution, but they rarely capture continuous maintenance at the granularity of package releases, where changes are bundled, shipped, and inherited by subsequent versions. We present SWE-Chain, a benchmark for evaluating agents on chained release-level package upgrades, where each transition builds on the agent's prior codebase. To produce upgrade specifications, we design a divide-and-conquer synthesis pipeline that aligns release notes with code diffs for each version transition, ensuring the requirements are grounded in actual code changes, informative to agents, and feasible to implement. SWE-Chain contains 12 upgrade chains across 9 real Python packages, with 155 version transitions and 1,660 grounded upgrade requirements. Across nine frontier agent-model configurations, agents achieve an average of 44.8% resolving, 65.4% precision, and 50.2% F1 under the Build+Fix regime, with Claude-Opus-4.7 (Claude Code) leading at 60.8% resolving, 80.6% precision, and 68.5% F1. These results show that SWE-Chain is both feasible and discriminative, and reveal that current agents still struggle to make correct upgrades across chained package releases without breaking existing functionality.

Out-of-distribution (OOD) detection is a critical component for ensuring the reliability of deep neural networks in safety-critical applications. In this work, we present a key empirical observation: for in-distribution (ID) samples, class-wise Mahalanobis distances exhibit a pronounced sharp minimum structure, where the distance to the nearest class is small while distances to all other classes remain large, resulting in high variance across classes. In contrast, OOD samples tend to exhibit a less pronounced sharp minimum structure, producing comparatively lower variance across classes. We further provide a theoretical analysis grounding this observation in Neural Collapse geometry: under relaxed Neural Collapse assumptions on within-class compactness and inter-class separation, ID samples are shown to structurally exhibit high class-wise distance variance, offering a theoretical basis for its use as an OOD score. Motivated by this observation and its theoretical backing, we propose MahaVar, a simple and effective post-hoc OOD detector that augments the Mahalanobis distance with a class-wise distance variance term. Following the OpenOOD v1.5 benchmark protocol, MahaVar achieves state-of-the-art performance on CIFAR-100 and ImageNet, with consistent improvements in both AUROC and FPR@95 over existing Mahalanobis-based methods across all benchmarks.

Quadruped robots are often designed with rigid feet to simplify control and maintain stable contact during locomotion. While this approach is straightforward, it limits the ability of the legs to absorb impact forces and reuse stored elastic energy, leading to higher energy expenditure during locomotion. To explore whether compliant feet can provide an advantage, we integrate foot compliance into a reinforcement learning (RL) locomotion controller and study its effect on walking efficiency. In simulation, we train eight policies corresponding to eight different spring stiffness values and then cross-evaluate their performance by measuring mechanical energy consumed per meter traveled. In experiments done on a developed quadruped, the energy consumption for the intermediate stiffness spring is lower by ~ 17% when compared to a very stiff or a very flexible spring incorporated in the feet, with similar trends appearing in the simulation results. These results indicate that selecting an appropriate foot compliance can improve locomotion efficiency without destabilizing the robot during motion.

The dominant discourse on AI limitations frames the boundary of AI capability as a divide between digital tasks (where AI excels) and physical tasks (where embodiment is required). We argue this framing misses the most consequential boundary: the one within digital tasks. We identify a class of tasks we call Metis AI, named for the Greek concept of metis (practical, contextual knowledge), that are performed entirely on computers yet resist reliable AI automation. These tasks are not computationally intractable; they are institutionally, socially, and normatively entangled in ways that defeat algorithmic approaches. We distinguish constitutive metis (knowledge destroyed by the act of formalization) from operational metis (system-specific familiarity that automation can progressively absorb), and propose five structural characteristics that define the Metis AI zone: consequential irreversibility, relational irreducibility, normative open texture, adversarial co-evolution, and accountability anchoring. We ground each in established theory from across the social sciences, philosophy, and humanitarian practice, argue that these characteristics are properties of the tasks themselves rather than limitations of current models, and show that the appropriate design response is not better automation but centaur architectures in which humans lead and AI supports.

Large-scale pretraining on Earth observation imagery has yielded powerful representations of the natural and built environment. However, most existing geospatial foundation models do not directly model the structured socioeconomic covariates typically stored in tabular form. This modality gap limits their ability to capture the complete total environment, which is critical for reasoning about complex environmental, social, and health-related outcomes. In this work, we propose GeoViSTA (Geospatial Vision-Tabular Transformer), a vision-tabular architecture that learns unified geospatial embeddings from co-registered gridded imagery and tabular data. GeoViSTA utilizes bilateral cross-attention to exchange spatial and semantic information across modalities, guided by a geography-aware attention mechanism that aligns continuous image patches with irregular census-tract tokens. We train GeoViSTA with a self-supervised joint masked-autoencoding objective, forcing it to recover missing image patches and tabular rows using local spatial context and cross-modal cues. Empirically, GeoViSTA's unified embeddings improve linear probing performance on high-impact downstream tasks, outperforming baselines in predicting disease-specific mortality and fire hazard frequency across held-out regions. These results demonstrate that jointly modeling the physical environment alongside structured socioeconomic context yields highly transferable representations for holistic geospatial inference.

Chaotic systems pose fundamental challenges for data-driven dynamics discovery, as small modeling errors lead to exponentially growing trajectory discrepancies. Since exact long-term prediction is unattainable, it is natural to ask what a good surrogate model for chaotic dynamics is. Prior work has largely focused either on reproducing the Jacobian of the underlying dynamics, which governs local expansion and contraction rates, or on training surrogate models that reproduce the ground-truth dynamics' long-term statistical behavior. In this work, we propose a new framework that aims to bridge these two paradigms by training surrogate dynamics models with accurate Jacobians and long-term statistical properties. Our method constructs a local covering of a chaotic attractor in phase space and analyzes the expansion and contraction of these coverings under the dynamics. The surrogate model is trained by minimizing the maximum mean discrepancy between the pushforward distributions of the coverings under the surrogate and ground-truth dynamics. Experiments show that our method significantly improves Jacobian accuracy while remaining competitive with state-of-the-art statistically accurate dynamics learning methods. Our code is fully available at https://anonymous.4open.science/r/neighborwatch.

While LLMs are increasingly used in commercial services, they pose privacy risks such as leakage of sensitive personally identifiable information (PII). For LLMs trained on multilingual corpora, Multilingual Machine Unlearning (MMU) aims to remove information across multiple languages. However, prior MMU evaluations fail to capture such cross-linguistic distribution of information, being largely limited to direct extensions of per-language evaluation protocols. To this end, we propose two metrics to evaluate the information spread across languages: the Knowledge Separability Score (KSS) and the Knowledge Persistence Score (KPS). KSS measures the overall unlearning quality across multiple languages, while KPS more specifically aims to assess consistent removal of information among different language pairs. We evaluated various unlearning methods in the multilingual setting with these metrics and conducted comprehensive analyses. Through our investigation, we provide insights into unique phenomena exclusive to MMU and offer a new perspective on MMU evaluation.

Memory-augmented LLM agents have advanced personalized recommendation, yet existing approaches universally adopt flat memory representations that conflate ephemeral signals with stable preferences, and none provides a complete lifecycle governing how memory should evolve. We propose MARS (Memory-Augmented Agentic Recommender System), a framework that treats recommendation as a partially observable problem and maintains a structured belief state that progressively abstracts noisy behavioral observations into a compact estimate of user preferences. MARS organizes this belief state into three tiers: event memory buffers raw signals, preference memory maintains fine-grained mutable chunks with explicit strength and evidence tracking, and profile memory distills all preferences into a coherent natural language narrative. A complete lifecycle of six operations -- extraction, reinforcement, weakening, consolidation, forgetting, and resynthesis -- is adaptively scheduled by an LLM-based planner rather than fixed-interval heuristics. Experiments on four InstructRec benchmark domains show that \ours achieves state-of-the-art performance with average improvements of 26.4% in HR@1 and 10.3% in NDCG@10 over the strongest baselines with further gains from agentic scheduling in evolving settings.

World models have emerged as a powerful paradigm for building interactive simulation environments, with recent video-based approaches demonstrating impressive progress in generating visually plausible dynamics. However, because these models typically infer dynamics from video and represent them in latent states, they do not explicitly enforce physical constraints. As a result, the generated video rollouts are not physically plausible, exhibiting unstable contacts, distorted shapes, or inconsistent motion. In this paper, we present an agentic framework constructing physics-based world models through executable simulation code. The framework coordinates planning, code generation, visual review, and physics analysis agents. The planning agent converts the natural language prompt into a structured scene plan, the code agent implements it as executable simulation code, and the visual review agent provide visual feedback while the physics analysis agent checks physical consistency. The code is iteratively revised based on the feedback until the simulation matches the prompt reqirements and physical constraints. Experimental results show that our framework outperforms advanced video-based models in physical accuracy, instruction fidelity and visual quality, which could be applied to various scenarios including driving simulation and embodied robot tasks.

Autonomous vehicles depend on online HD map construction to perceive lane boundaries, dividers, and pedestrian crossings -- safety-critical road elements that directly govern motion planning. While existing pixel perturbation attacks can disrupt the mapping, they can be neutralized by standard adversarial defenses. We present MIRAGE, a framework for systematic discovery of semantic attacks that bypass adversarial defenses and degrade mapping predictions by finding plausible environmental variation (e.g. shadows, wet roads). MIRAGE exploits the latent manifold of real-world data learned by diffusion models, and searches for semantically mutated scenes neighboring the ground truth with the same road topology yet mislead the mapping predictions. We evaluate MIRAGE on nuScenes and demonstrate two attacks: (1) boundary removal, suppressing 57.7% of detections and corrupting 96% of planned trajectories; and (2) boundary injection, the only method that successfully injects fictitious boundaries, while pixel PGD and AdvPatch fail entirely. Both attacks remain potent under various adversarial defenses. We use two independent VLM judges to quantify realism, where MIRAGE passes as realistic 80--84% of the time (vs. 97--99% for clean nuScenes), while AdvPatch only 0--9%. Our findings expose a categorical gap in current adversarial defenses: semantic-level perturbations that manifest as legitimate environmental variation are substantially harder to mitigate than pixel-level perturbations.

We pursue a vision for self-improving language models in which the model does not merely generate problems or traces to imitate, but constructs the environments that train it. In zero-data reasoning RL, this reframes self-improvement from a data-generation loop into an environment-construction loop, where each artifact is a reusable executable object that samples instances, computes references, and scores responses. Whether this vision sustains improvement hinges on a single property: the environments must exhibit stable solve--verify asymmetry, the model must be able to write an oracle once that it cannot reliably execute in natural language on fresh instances. This asymmetry takes two complementary forms. Some tasks are algorithmically hard to reason through but trivial as code: a dynamic program or graph traversal, compiled once, yields unboundedly many calibrated instances. Others are intrinsically hard to solve but easy to verify, like planted subset-sum or constraint satisfaction. Both create a durable gap between proposing and solving that the policy cannot close by gaming the verifier, and it is this gap that keeps reward informative as the learner improves. We instantiate this view in EvoEnv, a single-policy generator, solver method that synthesizes Python environments from ten seeds and admits them only after staged validation, semantic self-review, solver-relative difficulty calibration, and novelty checks. The strongest evidence comes from the already-strong regime: on Qwen3-4B-Thinking, fixed public-data RLVR and fixed hand-crafted environment RLVR reduce the average, while EvoEnv improves it from 72.4 to 74.8, a relative gain of 3.3%. Stable self-improvement, we suggest, depends not on producing more synthetic data, but on models learning to construct worlds whose difficulty stays structurally beyond their own reach.

Time series forecasting is not just numerical extrapolation, but often requires reasoning with unstructured contextual data such as news or events. While specialized Time Series Foundation Models (TSFMs) excel at forecasting based on numerical patterns, they remain unaware to real-world textual signals. Conversely, while LLMs are emerging as zero-shot forecasters, their performance remains uneven across domains and contextual grounding. To bridge this gap, we introduce Nexus, a multi-agent forecasting framework that decomposes prediction into specialized stages: isolating macro-level and micro-level temporal fluctuations, and integrating contextual information when available before synthesizing a final forecast. This decomposition enables Nexus to adapt from seasonal signals to volatile, event-driven information without relying on external statistical anchors or monolithic prompting. We show that current-generation LLMs possess substantially stronger intrinsic forecasting ability than previously recognized, depending critically on how numerical and contextual reasoning are organized. Evaluated on data strictly succeeding LLM knowledge cutoffs spanning Zillow real estate metrics and volatile stock market equities, Nexus consistently matches or outperforms state-of-the-art TSFMs and strong LLM baselines. Beyond numerical accuracy, Nexus produces high-quality reasoning traces that explicitly show the fundamental drivers behind each forecast. Our results establish that real-world forecasting is an agentic reasoning problem extending well beyond only sequence modeling.

We present Darwin Family, a framework for training-free evolutionary merging of large language models via gradient-free weight-space recombination. We ask whether frontier-level reasoning performance can be improved without additional training, by reorganizing latent capabilities already encoded in existing checkpoints. Darwin introduces three key ideas: (i) a 14-dimensional adaptive merge genome enabling fine-grained component- and block-level recombination; (ii) MRI-Trust Fusion, which adaptively balances diagnostic layer-importance signals with evolutionary search through a learnable trust parameter; and (iii) an Architecture Mapper that enables cross-architecture breeding between heterogeneous model families. Empirically, the flagship Darwin-27B-Opus achieves 86.9% on GPQA Diamond, ranking #6 among 1,252 evaluated models, and outperforming its fully trained foundation model without any gradient-based training. Across scales from 4B to 35B parameters, Darwin models consistently improve over their parents, support recursive multi-generation evolution, and enable a training-free evolutionary merge that combines Transformer- and Mamba-based components. Together, the Darwin Family demonstrates that diagnostic-guided evolutionary merging is a practical and reproducible alternative to costly post-training pipelines for reasoning-centric language models.

Recent advancements in generative AI facilitate large-scale synthetic data generation for model evaluation. However, without targeted approaches, these datasets often lack the sociotechnical nuance required for sensitive domains. We introduce NodeSynth, an evidence-grounded methodology that generates socially relevant synthetic queries by leveraging a fine-tuned taxonomy generator (TaG) anchored in real-world evidence. Evaluated against four mainstream LLMs (e.g., Claude 4.5 Haiku), NodeSynth elicited failure rates up to five times higher than human-authored benchmarks. Ablation studies confirm that our granular taxonomic expansion significantly drives these failure rates, while independent validation reveals critical deficiencies in prominent guard models (e.g., Llama-Guard-3). We open-source our end-to-end research prototype and datasets to enable scalable, high-stakes model evaluation and targeted safety interventions (https://github.com/google-research/nodesynth).

Psychological defense mechanisms (PDMs) are unconscious cognitive processes that modulate how individuals perceive and respond to emotional distress. Automatically classifying PDMs from text is clinically valuable but severely hindered by data scarcity and class imbalance, challenges which generative augmentation alone cannot resolve without psychological grounding. In this work, we address these challenges in the PsyDefDetect shared task (BioNLP@ACL 2026) by proposing a context-aware synthetic augmentation framework combined with a hybrid classification model. Our hybrid model integrates contextual language representations with basic clinical features, along with 150 annotated defense items. Experiments demonstrate that definition quality in prompting directly governs generation fidelity and downstream performance. Our method surpasses DMRS Co-Pilot, reaching an accuracy of 58.26% (+40.25%) and a macro-F1 of 24.62% (+15.99%), thereby establishing a strong baseline for psychologically grounded defense mechanism classification in low-resource settings. Source code is available at: https://github.com/htdgv/CASA-PDC.

Finding approximate equilibria for large-scale imperfect-information competitive games such as StarCraft, Dota, and CounterStrike remains computationally infeasible due to sparse rewards and challenging exploration over long horizons. In this paper, we propose a multi-agent starting-state sampling strategy designed to substantially accelerate online exploration in regularized policy-gradient game methods for two-player zero-sum (2p0s) games. Motivated by an assumption that offline demonstrations from skilled humans can provide good coverage of high-level strategies relevant to equilibrium play, we propose the initialization of reinforcement learning data collection at intermediate states sampled from offline data to facilitate exploration of strategically relevant subgames. Referring to this method as Data-Augmented Game Starts (DAGS), we perform experiments using synthetic datasets and analytically tractable, long-horizon control variants of two-player Kuhn Poker, Goofspiel, and a counterexample game designed to penalize biased beliefs over hidden information. Under fixed computational budgets, DAGS enables regularized policy gradient methods to achieve lower exploitability in games with significantly more challenging exploration. We show that augmenting starting state distributions when solving imperfect information games can lead to biased equilibria, and we provide a straightforward mitigation to this in the form of multi-task observation flags. Finally, we release a new set of benchmark environments that drastically increase exploration challenges and state counts in existing OpenSpiel games while keeping exploitability measurements analytically tractable.

Pattern discovery in data plays a crucial role across diverse domains, including healthcare, risk assessment, and machinery maintenance. In contrast to black-box deep learning models, symbolic rule discovery emerges as a key data mining task, generating human-interpretable rules that offer both transparency and intuitive explainability. This paper introduces the Optimal Pattern Detection Tree (OPDT), a rule-based machine learning model based on novel mixed-integer programming to discover a single optimal pattern in data through binary classification. To incorporate prior knowledge and compliance requirements, we further introduce the Branching Structure Constraints (BSC) framework, which enables decision makers to encode domain knowledge and constraints directly into the model. This optimization-based approach discovers a hidden underlying pattern in datasets, when it exists, by identifying an optimal rule that maximizes coverage while minimizing the false positive rate due to misclassification. Our computational experiments show that OPDT discovers a pattern with optimality guarantees on moderately sized datasets within reasonable runtime.

Zeroth-Order (ZO) optimization is pivotal for scenarios where backpropagation is unavailable, such as memory-constrained on-device learning and black-box optimization. However, existing methods face a stark trade-off: they are either sample-inefficient (e.g., standard finite differences) or suffer from high variance due to randomized estimation (e.g., random subspace methods). In this work, we propose Coherent Coordinate Descent (CoCD), a deterministic, sample-efficient, and budget-aware ZO optimizer. Theoretically, we formalize the notion of gradient coherence and demonstrate that CoCD is equivalent to Block Cyclic Coordinate Descent (BCCD) with ``warm starts,'' effectively converting historical (stale) gradients from a liability into a computational asset. This mechanism enables $O(1)$ query complexity per step while maintaining global descent directions. Furthermore, we derive error bounds revealing a counter-intuitive insight: larger finite-difference step sizes can induce an implicit smoothing effect on the optimization landscape by reducing the effective smoothness constant, thereby improving convergence stability. Experiments on MLP, CNN, and ResNet architectures (up to 270k parameters) demonstrate that CoCD significantly outperforms BCCD in terms of sample efficiency and convergence loss/accuracy, and exhibits superior stability over randomized ZO methods. Our results suggest that deterministic, structure-aware updates offer a superior alternative to randomization for lightweight ZO optimization.

Full-waveform inversion (FWI) estimates unknown parameters in the wave equation from limited boundary measurements. Recent advances in neural reparameterized FWI (NeurFWI) demonstrate that representing the parameters using a neural network can reduce the reliance on the high-quality initial model and wavefield data, at the cost of slow high-resolution convergence. However, its underlying theoretical mechanism remains unclear. In this study, we establish the neural sensitivity kernel (NSK) and the wave tangent kernel (WTK) to analyze their convergence behavior from both model and data domains. These theoretical frameworks show that the neural tangent kernel (NTK) induced by neural representation adaptively modulates the original sensitivity and wave tangent kernels. This modulation leads to several key outcomes, i.e., the spectral filtering effect, the gradient wavenumber modulation, and the wave frequency bias, connecting the convergence behavior of NeurFWI with the eigen-structures of NSK and WTK. Building on these insights, we propose several enhanced NeurFWI methods with tailored eigen-structures in NSK and WTK to improve inversion performances and efficiency. We numerically validate these theoretical claims and the proposed methods in seismic exploration, and firstly extend their application to medical imaging.

Continuous diffusion language models lag behind autoregressive transformers, partly because diffusion is applied in spaces poorly suited to language denoising and token recovery. We propose DiHAL, a geometry-guided diffusion-transformer hybrid that asks where diffusion should enter a pretrained transformer. DiHAL scores layers with geometry-based proxies, selects a diffusion-friendly hidden-state interface, and replaces the lower transformer prefix with a diffusion bridge while retaining the upper layers and original LM head. By reconstructing the selected-layer hidden state rather than tokens, DiHAL avoids direct continuous-to-discrete recovery. Experiments on 8B-scale backbones show that the geometry score predicts effective shallow insertion layers under a fixed bridge-training protocol and that hidden-state recovery improves over continuous diffusion baselines in a diagnostic comparison matching the diffusion/recovery training budget. These results suggest that hidden-state geometry helps identify where diffusion-based replacement is feasible inside pretrained language models.

Extending large language models (LLMs) to low-resource languages often incurs an "alignment tax": improvements in the target language come at the cost of catastrophic forgetting in general capabilities. We argue that this trade-off arises from the rigidity of supervised fine-tuning (SFT), which enforces token-level surface imitation on narrow and biased data distributions. To address this limitation, we propose a semantic-space alignment paradigm powered by Group Relative Policy Optimization (GRPO), where the model is optimized using embedding-level semantic rewards rather than likelihood maximization. This objective encourages meaning preservation through flexible realizations, enabling controlled updates that reduce destructive interference with pretrained knowledge. We evaluate our approach on Tibetan-Chinese machine translation and Tibetan headline generation. Experiments show that our method acquires low-resource capabilities while markedly mitigating alignment tax, preserving general competence more effectively than SFT. Despite producing less rigid surface overlap, semantic RL yields higher semantic quality and preference in open-ended generation, and few-shot transfer results indicate that it learns more transferable and robust representations under limited supervision. Overall, our study demonstrates that reinforcement learning with semantic rewards provides a safer and more reliable pathway for inclusive low-resource language expansion.

Recent tabular learning benchmarks increasingly show a tight performance cluster rather than a clear hierarchy among leading methods, spanning gradient boosted decision trees, attention-based architectures, and implicit ensembles such as TabM. As benchmark gains plateau, a complementary goal is to understand and control the mechanisms that make simple neural tabular models competitive. We propose LoMETab, a rank-$r$ generalization of multiplicative implicit ensembles. LoMETab lifts the rank-1 BatchEnsemble/TabM modulation to a rank-$r$ identity-residual Hadamard family by parameterizing each member weight as $W_k = W \odot (1 + A_kB_k^\top)$, where $W$ is shared and $(A_k, B_k)$ are member-specific low-rank factors. This exposes two practical diversity-control axes: the adapter rank $r$ and the initialization scale $σ_{\mathrm{init}}$, and we prove that for $r \ge 2$ this generalization strictly enlarges BatchEnsemble's hypothesis class. Empirically, we show that this added capacity manifests as measurable predictive diversity after training: on representative classification datasets, LoMETab sustains higher pairwise KL than an additive low-rank ablation, and $(r, σ_{\mathrm{init}})$ provides broad control over pairwise KL, varying by up to several orders of magnitude across configurations. The induced diversity is reflected in task-appropriate output-level measures: argmax disagreement for classification and ambiguity for regression, indicating that the control extends beyond pairwise KL to decision- and output-level member variation. Finally, experiments sweeping over adapter rank $r$ and initialization scale $σ_{\mathrm{init}}$ reveal that predictive performance is dataset-dependent over the $(r, σ_{\mathrm{init}})$ grid, supporting LoMETab as a controllable family of implicit ensembles rather than a fixed rank-1 construction.

Continual learning requires models to adapt to new data while preserving previously acquired knowledge. At its core, this challenge can be viewed as principled one-step adaptation: incorporating new information with minimal interference to existing representations. Most existing approaches address this challenge by modifying model parameters or architectures in a supervised, task-specific manner. However, the underlying issue is representational: tasks require distinct yet structured representations that can be selectively updated without disrupting representations, while structure should reflect intrinsic organization in the data rather than task boundaries. In sequential data, time-delayed dependencies provide a natural signal for uncovering this organization, revealing how fundamental representations give rise to more specific ones. Inspired by the modular organization of the human brain, we propose MoRe, a framework that identifies modularity in the representation itself rather than allocating it at the architectural level. MoRe decomposes knowledge into a hierarchy of fundamental and specific modules with identifiability guarantees, enabling principled module reuse, alignment, and expansion during adaptation while preserving old modules by construction. Experiments on synthetic benchmarks and real-world LLM activations demonstrate interpretable hierarchical structure, improved plasticity-stability trade-offs, suggesting MoRe as a principled foundation for continual adaptation

Context window efficiency is a practical constraint in large language model (LLM)-based developer tools. Paulsen [12] shows that all tested models degrade in accuracy well before their advertised context limits the Maximum Effective Context Window (MECW) which makes context construction a quality problem, not just a cost one. Modern software repositories routinely contain large non-code artifacts compiled datasets, binary model weights, minified JavaScript bundles, and gigabyte-scale log files that overflow the context window and push out task-relevant source code. We present a correctness-aware context hygiene framework: a pre-execution, size-based heuristic filter that intercepts repository scans before tokenization, using only OS-level stat() metadata with sub-millisecond overhead. Semantic retrieval approaches such as RepoCoder, GraphRAG, and AST-based chunking require index construction and query-time inference before any filtering decision is reached. Our framework, by contrast, requires no indexing and operates at <0.01 ms per file decision. Across 10 real open-source repositories (22,046 files, 5 languages), the proposed SizeFilter at θ=1 MB achieves 79.6% (\pm13.2%) mean token reduction at 0.30 ms overhead: the HybridFilter achieves 89.3% (\pm9.0%) the lowest variance of any filter evaluated. A token-density study across 2,688 files confirms a strong linear correlation (Pearson r=0.997, k=0.250 tokens/byte). A limited-scope evaluation (18 tasks, CodeLlama-7B-Instruct) yields 72% file-level accuracy under filtering versus 25% at baseline; hallucination frequency declines from 61% to 17%. All code and data are released for reproducibility.

Wearable devices capture physiological and behavioral data with increasing fidelity, but the psychological context shaping these outcomes is difficult to recover from sensor data alone, limiting passive sensing utility for digital health. We examined whether ultra-brief naturalistic concern text could serve as a scalable complement to passive sensing. In a year-long study of 458 university students (3,610 person-waves) tracked with Oura rings, participants responded bimonthly to an open-ended prompt about what concerned them most; responses had a median length of three words. We compared dictionary-based, general pretrained, and domain-adapted NLP approaches using within-person mixed-effects models across nine sleep and physical activity outcomes. Weeks dominated by academic concern framing were associated with lower physical activity; weeks characterized by emotional exhaustion language were associated with poorer sleep quality and lower heart rate variability. General pretrained embeddings outperformed domain-adapted models for most outcomes, with domain adaptation showing relative advantage for autonomic outcomes. Zero-shot classification of concern topics produced no significant associations, while affective dimensions across all three methods were consistently associated with outcomes, indicating emotional register rather than topical content carries the signal. These findings offer design guidance: ultra-brief affective prompts enrich the psychological interpretability of passive physiological data at minimal burden.

Vector quantization is a fundamental tool for compressing high-dimensional embeddings, yet existing multi-codebook methods rely on static codebooks that limit expressiveness under heterogeneous data geometry. While recent dynamic quantizers like QINCo adapt codebooks to individual inputs and improve expressiveness, their strict sequential dependencies create decoding bottlenecks. We propose Residual Quantization via Mixture of Experts (RQ-MoE), a framework combining a two-level MoE with dual-stream quantization to enable input-dependent codebook adaptation for efficient vector quantization. RQ-MoE enables dynamic codebook construction and decouples instruction from quantization, facilitating parallel decoding. Theoretically, we show that standard Residual Quantization and QINCo can be recovered as constrained special cases of RQ-MoE, and derive a guideline for setting expert dimensionality in RQ-MoE. Extensive experiments show that RQ-MoE achieves state-of-the-art or on-par performance in reconstruction and retrieval, while providing 6x-14x faster decoding than prior vector quantization methods. The implementation is available at https://github.com/KDEGroup/RQ-MoE.

Language models often generate long chain-of-thought traces, but it remains unclear how much of this reasoning is necessary for preserving the final prediction. We study this through the lens of overcomplete reasoning traces: generated traces that contain more intermediate steps than are needed to support the model's answer. We define the minimal core as the smallest subset of steps that preserves either the final answer or predictive distribution, and introduce metrics for compression ratio, redundancy mass, step necessity, and necessity concentration. Across six deliberative reasoning benchmarks spanning arithmetic, competition mathematics, expert scientific reasoning, and commonsense multi-hop QA, we find substantial overcompleteness: on average, 46% of steps are removable under greedy minimal-core extraction while preserving the original answer in 86% of cases. We also find that predictive support is concentrated: the top three steps account for 65% of measured necessity mass on average. Beyond compression, minimal cores expose a cleaner geometry of reasoning: compared with full traces, they improve correct-incorrect trace separation by 11 points, reduce estimated intrinsic dimensionality by 34%, and transfer across model families with 85% off-diagonal answer retention. Theoretically, we establish existence of minimal sufficient subsets, local irreducibility guarantees for greedy elimination, and certificates of overcompleteness and sparse necessity. Together, these results suggest that full reasoning traces are often verbose and overcomplete, while minimal cores isolate the effective support underlying language-model predictions.

As AI agents improve, the central question is no longer whether they can solve isolated well-defined financial tasks, but whether they can reliably carry out financial professional work. Existing financial benchmarks offer only a partial view of this ability, as they primarily evaluate static competencies such as question answering, retrieval, summarization, and classification. We introduce Herculean, the first skilled benchmark for agentic financial intelligence spanning four representative workflows, including Trading, Hedging, Market Insights, and Auditing. Each workflow is instantiated as a standardized MCP-based skill environment with its own tools, interaction dynamics, constraints, and success criteria, enabling consistent end-to-end assessment of heterogeneous agent systems. Across frontier agents, we find agents perform relatively well on Trading and Market Insights, but struggle substantially on Hedging and Auditing, where long-horizon coordination, state consistency, and structured verification are critical. Overall, our results point to a key gap in current agents in turning financial reasoning into dependable workflow execution in high-stakes financial workflows.

We present a new computational framework for detecting and structuring manipulative political narratives. A task that became more important due to the shift of political discussions to social media. One of the primary challenges thereby is differentiating between manipulative political narratives and legitimate critiques. Some posts may also reframe actual events within a manipulative context. To achieve good clustering results, we filter manipulative posts beforehand using a detailed few-shot prompt that combines documented campaign narratives with legitimate criticisms to differentiate them. This prompt enables a reasoning model to assign labels, retaining only manipulative narrative posts for further processing. The remaining posts are subsequently embedded and dimensionality-reduced using UMAP, before HDBSCAN is applied to uncover narrative groups. A key advantage of this unsupervised approach is its independence from a predefined list of target categories, enabling it to uncover new narrative clusters. Finally, a reasoning model is employed to uncover the narrative behind each cluster. This approach, applied to over 1.2 million social media posts, effectively identified 41 distinct manipulative narrative clusters by integrating prompt-based filtering with unsupervised clustering.

Elections represent a crucial milestone in a nation's ongoing development. To better understand the political rhetoric from various movements, ranging from left to right, we propose a transformer-based model capable of projecting the political orientation of a text on a continuous left-to-right spectrum, represented by a normalized scalar d between -1 and 1. This approach enables analysts to focus on specific segments of the political landscape, such as conservatives, while excluding liberal and far-right movements. Such a task can only be achieved with multiclass classifiers, provided that the desired orientation is incorporated within one of their predefined classes. To determine the most suitable foundation model among 13 candidate transformers for this task, we constructed four distinct corpora. One corpus comprised annotated plenary notes from the German Bundestag, while another was based on an official online decision-making tool, Wahl-O-Mat. The third corpus consisted of articles from 33 newspapers, each identified by its political orientation, and the fourth included 535,200 tweets from 597 members of the 20th and 21st German Bundestag. To mitigate overfitting, we used two distinct corpora for training and two for testing, respectively. For in-domain performance, DeBERTa-large achieved the highest F1 score F1=0.844 as well as for the X (Twitter) out-of-domain test ACC=0.864. Regarding the newspaper out-of-domain test, Gemma2-2B excelled (MAE = 0.172). This study demonstrates that transformer models can recognize political framing in German news at the level of public opinion polls. Our findings suggest that both the model architecture and the availability of domain-specific training data can be as influential as model size for estimating political bias. We discuss methodological limitations and outline directions for improving the robustness of bias measurement.

We present a physics-informed framework for system identification based on randomized stable atomic features. Impulse responses are represented as random superpositions of stable atoms, namely damped complex exponentials associated with poles sampled inside a prescribed disk. Identification is then cast as a convex regularized least-squares problem with optional linear, second-order-cone, and KYP constraints. The approach generalizes random Fourier and random Laplace features to the damped, nonstationary regime relevant to engineering systems while retaining modal interpretability and scalable finite-dimensional computation. The main analytic point is an operator-theoretic Disk-Bochner viewpoint: positive measures over stable poles generate positive-definite kernels with a radius-dependent shift defect, while a converse scalar disk moment representation for an arbitrary kernel is characterized by subnormality of the canonical shift. We prove this statement, establish an RKHS-to-l1 embedding, show that sampled poles induce a valid finite atomic gauge, discuss random-feature convergence, and state sparse-recovery guarantees conditionally on the restricted-eigenvalue properties of the realized disk-Vandermonde or input-output design matrix. We also connect the normalized transfer function problem to Nevanlinna-Pick interpolation and LFT set-membership. The framework directly encodes stability margins, modal localization, DC-gain bounds, monotonicity, passivity, relative degree, settling-time targets, and time/frequency-domain error bounds. Numerical comparisons illustrate how physically meaningful priors can compensate for poor excitation and improve constrained impulse-response recovery in an under-informative data setting.

Multi-task reinforcement learning (MTRL) aims to train a single agent to efficiently optimize performance across multiple tasks simultaneously. However, jointly optimizing all tasks often yields imbalanced learning: agents quickly solve easy tasks but learn slowly on harder ones. While prior work primarily attributes this imbalance to conflicting task gradients and proposes gradient manipulation or specialized architectures to address it, we instead focus on a distinct and under-explored challenge: imbalanced data allocation. Standard MTRL allocates an equal number of environment interactions to each task, which over-allocates data to easy tasks that require relatively few interactions to solve and under-allocates data to hard tasks that require substantially more experience to solve. To address this challenge, we introduce Distributionally Robust Adaptive Task Sampling (DRATS), an algorithm that adaptively prioritizes sampling tasks furthest from being solved. We derive DRATS by formalizing MTRL as a feasibility problem from which we derive a minimax objective for minimizing the worst-case return gap, the difference between a desired target return and the agent's return on a task. In benchmarks like MetaWorld-MT10 and MT50, DRATS improves data efficiency and increases worst-task performance compared to existing task sampling algorithms.

We introduce Exemplar Partitioning (EP), an unsupervised method for constructing interpretable feature dictionaries from large language model activations with $\sim 10^{3}\times$ fewer tokens than comparable sparse autoencoders (SAEs). An EP dictionary is a Voronoi partition of activation space, built by leader-clustering streamed activations within a distance threshold. Each region is anchored by an observed exemplar that serves as both its membership criterion and intervention direction; dictionary size is not prespecified, but determined by the activation geometry at that threshold. Because exemplars are observed rather than learned, dictionaries built from the same data stream are directly comparable across layers, models, and training checkpoints. We characterise EP as an interpretability object via targeted demonstrations of properties newly accessible through this construction, plus one head-to-head benchmark. In Gemma-2-2B, EP dictionary regions are interpretable and support causal interventions: refusal in instruction-tuned Gemma concentrates in a region whose exemplar ablation can collapse held-out refusal. Cross-checkpoint matching between base and instruction-tuned dictionaries separates the directions preserved through finetuning from those introduced by it. EP regions and Gemma Scope SAE features decompose activation space differently but agree on a shared core: $\sim 20\%$ of EP regions match an SAE feature at $F_{1} > 0.5$, and EP one-hot probes retain $\sim 97\%$ of raw-activation probe accuracy at $\ell_{0} = 1$. Nearest-exemplar distance provides a free out-of-distribution signal at inference. On AxBench latent concept detection at Gemma-2-2B-it L20, EP at $p_{1}$ reaches mean AUROC $0.881$, $+0.126$ over the canonical GemmaScope SAE leaderboard entry and within $0.030$ of SAE-A's $0.911$, at $\sim 10^{3}\times$ less build compute.

Generative modeling has emerged as a promising approach for crystal structure discovery. However, existing LLM-based generative models struggle with low-level atomic precision, while diffusion-based methods fall short in integrating high-level scientific knowledge. As a result, generated structures are often invalid, unstable, or do not possess desirable properties. To address this gap, we propose CrystalReasoner (\method), an end-to-end LLM framework that generates crystal structures from natural language instructions through reasoning and alignment. \method introduces physical priors as thinking tokens, which include crystallographic symmetry, local coordination environments and predicted physical properties before generating atomic coordinates. This bridges the gap between natural language and 3D structures. \method then employs reinforcement learning (RL) with a multi-objective, dense reward function to align generation with physical validity, chemical consistency, and thermodynamic stability. For property-conditioned tasks, we design task-specific reward functions and train specialized models for discrete constraints (e.g., space group) and continuous properties (e.g., elasticity, thermal expansion). Empirical results demonstrate that compared to prior works and baselines without thinking traces or RL, \method obtains better performance on diverse metrics, triples S.U.N. ratio, and achieves better performance for property conditioned generation. \method also exhibits adaptive reasoning, increasing reasoning lengths as the number of atoms increases. Our work demonstrates the potential of leveraging thinking traces and RL for generating valid, stable, and property-conditioned crystal structures. Please see our work at https://crystalreasoner.github.io/ .

Nearest-neighbor methods are fundamental to classical and modern machine learning, yet their geometric properties are typically analyzed under independent sampling. In this paper, we study the nearest-neighbor radii under dependent sampling. We consider strong mixing dependent observations and ask whether dependence changes the scale of nearest-neighbor neighborhoods. We establish distribution-free almost sure convergence under polynomial mixing and sharp non-asymptotic moment bounds under geometric mixing. The moment bounds depend on the local intrinsic dimension rather than the ambient dimension, making the results applicable to high-dimensional data concentrated near lower-dimensional manifolds. Synthetic experiments and real-world time-series benchmarks support the theory, showing that nearest-neighbor geometry remains informative under dependence sampling.

Modern edge devices increasingly rely on neural networks for intelligent applications. However, conventional digital computing-based edge inference requires substantial memory and energy consumption. In analog radio frequency (RF) computing, a base station (BS) encodes the weights of the neural networks and broadcasts the RF waveforms to the clients. Each client reuses its passive mixer to multiply the received weight-encoded waveform with a locally generated input-encoded waveform. This enables wireless receivers to perform the matrix-vector multiplications (MVMs) that account for most of the computation burden in edge inference with ultra-low energy consumption. Unlike conventional downlink transmissions which are optimized for communications, analog RF computing requires a computing-centric physical layer that controls both the analog MVM accuracy and the energy consumption for inference. Motivated by this, in this paper, we propose a physical layer design framework for analog RF computing in MU-MIMO wireless systems. We derive tractable models for computing accuracy and energy consumption for inference, formulate a joint BS beamforming and client-side scaling problem subject to computing accuracy, transmit power, and hardware constraints, and develop a low-complexity algorithm to solve the non-convex problem. The proposed design provides client- and layer-specific accuracy control for both uniform- and mixed-precision inference. Simulations under 3GPP specifications show that analog RF computing can significantly reduce client-side energy consumption by nearly two orders of magnitude compared to digital computing, while mixed-precision inference requires even lower energy consumption than uniform-precision inference. Overall, these results establish analog RF computing over wireless networks as a promising paradigm for energy-efficient edge inference.

Drug-drug interaction (DDI) prediction is a critical task in computational biomedicine, as adverse interactions between co-administered drugs can cause severe side effects and clinical risks. A key challenge is unseen-drug generalization, where interactions must be predicted for drugs not observed during training. Although multimodal DDI models exploit diverse drug-related information, their fusion mechanisms are often tied to specific prediction architectures, limiting their reuse across models. To address this, we propose AIM-DDI, an architecture-independent multimodal integration module that represents heterogeneous modality information as tokens in a shared latent space. By modeling dependencies across modality tokens through a unified fusion module, AIM-DDI enables model-agnostic integration of structural, chemical, and semantic drug signals across different DDI prediction architectures. Extensive evaluations across diverse DDI models and DrugBank-based settings show that AIM-DDI consistently improves prediction performance, with the strongest gains under the most challenging both-unseen setting where neither drug in a test pair is observed during training. These results suggest that treating multimodal integration as a reusable module, rather than a model-specific fusion component, is an effective strategy for robust unseen-drug DDI prediction.

We investigate the temporal concatenation of sub-policies in Markov Decision Processes (MDP) with time-varying reward functions. We introduce General Dijkstra Search (GDS), and prove that globally optimal goal-reaching policies can be recovered through temporal composition of intermediate optimal sub-policies. Motivated by the "search, select, update" principle underlying GDS, we propose Dynamic Latent Routing (DLR), a language-model post-training method that jointly learns discrete latent codes, routing policies, and model parameters through dynamic search in a single training stage. In low-data fine-tuning settings, DLR matches or outperforms supervised fine-tuning across four datasets and six models, achieving a mean gain of +6.6 percentage points, while prior discrete-latent baselines consistently underperform SFT. Mechanistic analyses and targeted code ablations show that DLR learns structured routing behaviors with distinct causal roles.

Language agents are increasingly deployed in complex professional workflows, with tutoring emerging as a particularly high-stakes capability that remains largely unmeasured in existing benchmarks. Effective tutor agents require more than producing correct answers or executing accurate tool calls: a robust tutor must diagnose learner state, adapt support over time, make pedagogically justified decisions grounded in educational evidence, and execute interventions within realistic learning-management systems. We introduce EduAgentBench, a source-grounded benchmark for holistically evaluating tutor agents across the full scope of teaching work. It contains 150 quality-controlled tasks across three capability surfaces: professional pedagogical judgment, situated multi-turn tutoring, and Canvas-style teaching workflow completion. Tasks are constructed through a pedagogical-insight-driven pipeline and evaluated with complementary verification signals and human review. Across a comprehensive evaluation of frontier models, our findings reveal that current models are generally capable of bounded pedagogical judgment, but still fall short of professional teaching standards in situated tutoring and autonomous teaching-workflow execution. To our knowledge, EduAgentBench is the first theory-grounded and realistic benchmark for evaluating the holistic teaching capability of tutor agents, providing a measurement foundation for developing future tutor agents that can support realistic teaching work.

Predictive maintenance in complex systems is often complicated by the heterogeneity and redundancy of monitored variables,which can obscure fault-relevant information and reduce model interpretability. This work proposes a semantic feature segmentation framework that decomposes the monitored feature space into a canonical component,expected to retain the dominant predictive information, and a residual component containing structurally peripheral signals. The segmentation is defined through domain informed criteria and sets up monitoring variables into functional groups reflecting operational mechanisms such as throughput,latency,pressure,network activity,and structural state. To evaluate the effectiveness of this decomposition, we adopt a predictive perspective in which expected predictive risk is used as an operational proxy for task-relevant information. Experimental results obtained through time-aware cross-validation show that the canonical space consistently achieves lower predictive risk than the residual space across multiple temporal configurations, indicating that the semantic segmentation concentrates the most relevant information for fault anticipation. In addition, the canonical segments exhibit significantly stronger intra-segment coherence than inter-segment dependence, and this structural organization remains stable after redundancy reduction. When compared with the full feature space and with a Principal Component Analysis (PCA) representation, the canonical space carries out comparable predictive performance and furthermore preserves the semantic meaning of the original variables. These findings suggest that semantic feature segmentation provides an interpretable and information-preserving decomposition of monitoring signals, enabling competitive predictive performance without sacrificing the operational interpretability required in predictive maintenance applications.

Extreme precipitation causes severe societal and economic damage, and weather control has long been discussed as a potential mitigation strategy. However, to the best of our knowledge, perturbation-based interventions for weather control using data-driven weather forecasting models have not yet been explored. While adversarial attacks also generate perturbations that alter forecasts, they aim to exploit model artifacts and do not account for physical plausibility. In this paper, we propose a gradient-based guidance framework for precipitation-reduction interventions through diffusion sampling in diffusion-based weather forecasting models. Instead of directly perturbing atmospheric states, our method steers the diffusion sampling trajectory, enabling precipitation reduction while maintaining consistency with the atmospheric distribution. To assess physical plausibility, we evaluate from three perspectives: (i) vertical and variable-wise perturbation profiles, (ii) latent-space trajectory deviation, and (iii) cross-model transferability. Experiments on extreme precipitation events from WeatherBench2 demonstrate that our method achieves effective precipitation reduction while yielding more physically plausible interventions than adversarial perturbations.

The growing adoption of AI agents and the Model Context Protocol (MCP) has motivated organizations to expose existing REST APIs as agent-consumable tools. In our industrial context, this initiative targeted an ecosystem of 16 production APIs comprising approximately 600 endpoints. Although these APIs were stable and widely used within a microservice architecture, early proof-of-concept experiments revealed systematic failures in task planning, tool selection, and payload construction when accessed through MCP-based agents. Rather than attributing these failures to model limitations alone, we conducted an ecosystem-scale empirical assessment of the underlying OpenAPI documentation. We developed Hermes, a multi-agent LLM-based system that detects documentation and REST-related smells at the endpoint level and generates explainable diagnostic reports. The large-scale evaluation identified 2,450 smells across 600 endpoints, with deficiencies present in all analyzed operations. Practitioner validation confirmed high agreement with the detected issues while also revealing contextual trade-offs in remediation decisions. The findings suggested that structural validity within microservice environments does not guarantee semantic readiness for agent-based consumption. Based on this evidence, the organization revised its adoption strategy, prioritizing selective endpoint adaptation, redefining documentation standards, and integrating automated documentation assessment into API governance workflows. This case illustrates how systematic artifact-level evaluation can function as a strategic decision-support mechanism, reducing technological risk and guiding evidence-based AI adoption in industrial software ecosystems.

Test-Time Scaling (TTS), which samples multiple candidate actions and ranks them via a Critic Model, has emerged as a promising paradigm for generalist GUI agents. Its efficacy thus hinges on the critic's fine-grained ranking ability. However, existing GUI critic models uniformly adopt binary classification. Our motivational analysis of these models exposes a severe entanglement: scores for valid actions and plausible-but-invalid distractors become indistinguishable. We attribute this failure to two structural defects: Affordance Collapse--the hierarchical affordance space is compressed into 0/1 labels; and Noise Sensitivity--binary objectives overfit to noisy decision boundaries. To resolve this, we introduce BBCritic (Beyond-Binary Critic), a paradigm shift grounded in the Functional Equivalence Hypothesis. Through two-stage contrastive learning, BBCritic aligns instructions and actions in a shared Affordance Space, recovering the hierarchical structure that binary supervision flattens. We also present BBBench (Beyond-Binary Bench), the first GUI critic benchmark that pairs a dense action space with a hierarchical four-level taxonomy, enabling fine-grained ranking evaluation. Experimental results show that BBCritic-3B, trained without any extra annotation, outperforms 7B-parameter SOTA binary models. It demonstrates strong zero-shot transferability across platforms and tasks, supporting our methodological view: GUI critique is fundamentally a metric-learning problem, not a classification one.

Machine unlearning in Vision-Language Models (VLMs) is typically performed at the image or instance level, making it difficult to precisely remove target knowledge without affecting unrelated semantics. This issue is especially pronounced since a single image often contains multiple entangled concepts, including both target concepts to be forgotten and contextual information that should be preserved. In this paper, we propose an interpretable concept-level unlearning framework for VLMs, which constructs a compact task-specific concept vocabulary from the forgetting set using a multimodal large language model. In addition to modality alignment, visual representations are decomposed into sparse, nonnegative combinations of semantic concepts, providing an explicit interface for fine-grained knowledge manipulation. Based on this decomposition, our method formulates unlearning as concept-level optimization, where target concepts are selectively suppressed while intra-instance non-target semantics and global cross-modal knowledge are preserved. Extensive experiments across both in-domain and out-of-domain forgetting settings demonstrate that our method enables more comprehensive target forgetting, better preserves non-target knowledge within the same image, and maintains competitive model utility compared with existing VLM unlearning methods.

Discrete diffusion language models improve generation efficiency through parallel token prediction, but standard $X_0$ prediction methods introduce factorization errors by approximating the clean token posterior with independent token-wise distributions. This paper proposes Factorization-Error-Free Discrete Diffusion Language Modeling (FeF-DLLM), which replaces independent clean-token prediction with an exact prefix-conditioned factorization of the clean posterior to better preserve token dependencies. To reduce the sequential cost introduced by prefix conditioning, FeF-DLLM further incorporates speculative decoding within diffusion denoising, accelerating inference while maintaining the parallel prediction and re-masking properties of DLLMs. Theoretically, we prove that FeF-DLLM generates from the true joint distribution and derive its expected acceleration ratio. Experiments on GSM8K, MATH, HumanEval, and MBPP demonstrate that our method improves accuracy by an average of 5.04 percentage points while achieving an average inference speedup of $3.86\times$.

Compositional generalization in sequential decision-making requires identifying which parts of prior rollouts remain useful for new tasks. Existing methods reuse skills or predictive models, but often overlook rich local transition geometry and dynamics. We propose Matrix-Space Reinforcement Learning (MSRL), a geometric abstraction that represents trajectory segments through positive semidefinite matrix descriptors aggregating first- and second-order statistics of lifted one-step transitions. These descriptors expose shared hidden structure, support algebraic composition in an abstract matrix space, and reveal opportunities for transfer. We prove that the descriptor is well defined up to coordinate gauge, complete for the induced low-order additive signal class, additive under valid segment composition, and minimally sufficient among admissible additive descriptors. We further show that conditioning value functions on the trajectory-segment matrix yields a first-order smooth approximation of action values, enabling source-learned matrix-to-value mappings to bootstrap learning in new tasks. MSRL is plug-in compatible with standard model-free and model-based methods, while obstruction filtering rejects implausible compositions. Empirically, MSRL achieves the best average finite-budget target AUC of 0.73, outperforming MSRL from scratch (0.65), TD-MPC-PT+FT (0.63), and TD-MPC (0.57).

Domain adaptation faces a fundamental paradox in the cold-start regime. When target data is scarce, statistical methods fail to distinguish relevant source domains from irrelevant ones, which often leads to negative transfer. In this paper, we address this challenge by leveraging expert textual descriptions of the target domain, a resource that is often available but overlooked. We propose a probabilistic framework that translates these semantic descriptions into a choice model, namely a Language-Induced Prior (LIP), that learns the preferences from a pretrained Large Language Model (LLM). The LIP is then integrated into an Expectation-Maximization algorithm to identify source relevance. Methodologically, this framework is compatible with any parametric model where a likelihood is available. It allows the LIP to guide the selection of sources when target signals are weak, while gradually refining these choices as samples accumulate. Theoretically, we prove that the estimator roughly matches an oracle cold-start MSE under a correct prior, while remaining asymptotically consistent regardless of the quality of the LIP. Empirically, we validated the framework on a descriptive (Gaussian estimation), a predictive (C-MAPSS dataset), and a prescriptive task (MuJoCo hopper).

We study reinforcement learning in hybrid discrete-continuous action spaces, such as settings where the discrete component selects a regime (or index) and the continuous component optimizes within it -- a structure common in robotics, control, and operations problems. Standard model-free policy gradient methods rely on score-function (SF) estimators and suffer from severe credit-assignment issues in high-dimensional settings, leading to poor gradient quality. On the other hand, differentiable simulation largely sidesteps these issues by backpropagating through a simulator, but the presence of discrete actions or non-smooth dynamics yields biased or uninformative gradients. To address this, we propose Hybrid Policy Optimization (HPO), which backpropagates through the simulator wherever smoothness permits, using a mixed gradient estimator that combines pathwise and SF gradients while maintaining unbiasedness. We also show how problems with action discontinuities can be reformulated in hybrid form, further broadening its applicability. Empirically, HPO substantially outperforms PPO on inventory control and switched linear-quadratic regulator problems, with performance gaps increasing as the continuous action dimension grows. Finally, we characterize the structure of the mixed gradient, showing that its cross term -- which captures how continuous actions influence future discrete decisions -- becomes negligible near a discrete best response, thereby enabling approximate decentralized updates of the continuous and discrete components and reducing variance near optimality. All resources are available at github.com/MatiasAlvo/hybrid-rl.

Formal verification of transformers has become increasingly important due to their widespread deployment in safety-critical applications. Compared to classic neural networks, the inferences of transformers involve highly complex computations, such as dot products in self-attention layers, rendering their verification extremely difficult. Existing approaches explored over-approximation methods by constructing convex constraints to bound the output ranges of transformers, which can achieve high efficiency. However, they may sacrifice verification precision, and consequently introduce significant approximation error that leads to frequent occurrences of false alarms. In this paper, we propose a transformer verification approach that can achieve improved precision. At the core of our approach is a novel usage of ReLU, by which we represent a precise but non-linear bound for dot products such that we can further exploit the rich body of literature for convex relaxation of ReLU to derive precise bounds. We extend two classic approaches to the context of transformers, a rule-based one and an optimization-based one, resulting in two new frameworks for efficient and precise verification. We evaluate our approaches on different model architectures and robustness properties derived from two datasets about sentiment analysis, and compare with the state-of-the-art baseline approach. Compared to the baseline, our approach can achieve significant precision improvement for most of the verification tasks with acceptable compromise of efficiency, which demonstrates the effectiveness of our approach.

KV-cache compression at small budgets is a crowded design space spanning cache representation, head-wise routing, compression cadence, decoding behavior, and within-budget scoring. We study seven mechanisms across these five families under matched mean cache on long-form mathematical reasoning (MATH-500~\cite{hendrycks2021math}) with two distilled-reasoning models (Qwen-7B and Llama-8B variants of DeepSeek-R1-Distill~\cite{deepseek2025r1}) at budgets $b \in \{64, 128\}$. All seven were rejected. We then propose $α$, a one-function modification to the TriAttention~\cite{mao2026triattention} retention scorer that replaces argmax-top-$k$ with greedy facility-location-inspired selection under a V-space redundancy penalty controlled by a single weight $λ$. A pre-registered protocol tunes $λ$ on a frozen development split and confirms on a disjoint held-out split; with $λ= 0.5$, $α$ clears Bonferroni on two of the four (model, budget) cells (Qwen $b{=}128$ and Llama $b{=}64$), no cell is significantly negative, and the pre-registered Branch~A triggers. The finding is asymmetric: a minimal scoring modification beat heavier structural redesigns in this regime, and the combined matched-memory, sympy-graded, held-out confirmation protocol is the evidence standard that made the asymmetry visible.

The rapid advancement of Large Vision-Language Models (LVLMs) is increasingly accompanied by unauthorized scraping and training on multimodal web data, posing severe copyright and privacy risks to data owners. Existing countermeasures, such as machine unlearning and watermarks, are inherent post-hoc approaches that act only after intellectual property infringement has already occurred. In this work, we propose MMGuard to empower data owners to proactively protect their multimodal data against unauthorized LVLM fine-tuning. MMGuard generates unlearnable examples by injecting human-imperceptible perturbations that actively exploit the learning dynamics of LVLMs. By minimizing the training loss, the perturbation creates an optimization shortcut, causing the model to overfit to the noise and thereby degrading downstream performance when the perturbation is absent during inference. To further strengthen this defense, MMGuard introduces a cross-modal binding disruption, strategically shifting LVLM attention to enforce a spurious correlation between the noise and the training target with theoretical guarantees. Enhanced by an ensemble learning strategy for cross-model transferability, MMGuard is evaluated against nine open-source LVLMs across six datasets. Our comprehensive results demonstrate effective, stealthy, and robust protection under white-box, gray-box, and black-box threat models, establishing a mechanistic advantage in proactively defending against aggressive fine-tuning exploitation.

ReAct has become the default architecture across LLM agents, and many existing web agents follow this paradigm. We argue that it is the wrong default for web agents. Instead, web agents should default to plan-then-execute: commit to a task-specific program before observing runtime web content, then execute it. The reason is that web content mixes inputs from many parties. An e-commerce product page may combine a seller's listing, customer reviews and sponsored advertisements. Under ReAct, all of this content flows into the model when deciding on the next action, creating a direct path for prompt injections to steer the agent's control flow. Plan-then-execute changes this boundary: untrusted data may influence values or branches inside a predefined execution graph, but it cannot redefine the user task or cause the model to synthesize new actions at runtime. We analyze WebArena, a popular web agent benchmark, and find that all tasks are compatible with plan-then-execute, while 80% can be completed with a purely programmatic plan, without any runtime LLM subroutine. We identify the main barrier to adopting plan-then-execute on the web: For it to work well, tools must map cleanly to semantic actions, with effects known before execution, so agents have enough information to plan. The web does not naturally expose that interface. Browser tools such as click, type, and scroll have page-dependent meanings. Planning at this layer is near-sighted: the agent can only see actions on the current page, and later actions appear only after it acts. Closing this gap requires typed interfaces that turn website interactions from clicks and keystrokes to task-level operations. This is an infrastructure problem, not a modeling problem. Web tasks do not need reactivity by default; they need typed, complete, auditable website APIs.

Mixture-of-Experts (MoE) models scale capacity by combining specialized experts, but most existing approaches assume centralized access to training data. In practice, data are distributed across clients and cannot be shared due to privacy constraints, making unified MoE training challenging. We propose MetaMoE, a privacy-preserving framework that unifies independently trained, domain-specialized experts into a single MoE using public proxy data as surrogates for inaccessible private data. Central to MetaMoE is diversity-aware proxy selection, which selects client-domain-relevant and diverse samples from public data to effectively approximate private data distributions and supervise router learning. These proxies are further used to align expert training, improving expert coordination at unification time, while a context-aware router enhances expert selection across heterogeneous inputs. Experiments on computer vision and natural language processing benchmarks demonstrate that MetaMoE consistently outperforms recent privacy-preserving MoE unification methods. Code is available at https://github.com/ws-jiang/MetaMoE.

Data assimilation (DA) estimates the state of an evolving dynamical system from noisy, partial observations, and is widely used in scientific simulation as well as weather and climate science. In practice, filtering methods rely on frame-to-frame transition models. However, these models are fragile when observations are non-Markovian (when they form only a partial slice of a higher-dimensional latent state as in real-world weather data): they tend to accumulate errors over long horizons. At the same time, learned DA methods typically commit to a single regime, either filtering (nowcasting, real-time forecasting) or smoothing (retrospective reanalysis), which splits what should be a shared prior across application-specific pipelines. To address both issues, we introduce ForcingDAS, a unified and robust DA framework. Built on Diffusion Forcing with an independent noise level assigned to each frame, ForcingDAS learns a joint-trajectory prior instead of frame-to-frame transitions. This allows it to capture long-horizon temporal dependencies and reduce error accumulation. In addition, the same trained model spans the full filtering to smoothing spectrum at inference time. Specifically, nowcasting, fixed-lag smoothing, and batch reanalysis are selected through the inference schedule alone, without retraining. We evaluate ForcingDAS on 2D Navier-Stokes vorticity, precipitation nowcasting, and global atmospheric state estimation. Across all settings, a single model is competitive with or outperforms both learned and classical baselines that are specialized for individual regimes, with the largest gains observed on real-world weather benchmarks.

Comparative evaluation of multiple dynamic treatment policies is essential for healthcare and policy decisions, yet conventional longitudinal causal inference methods estimate each in isolation, preventing information sharing across counterfactuals. We demonstrate that this separate estimation paradigm induces a structurally uncontrolled second-order bias, inflating finite-sample variance even after standard debiasing with longitudinal targeted maximum likelihood estimation(LTMLE). To address this, we propose a policy-aware reparameterization of Iterative Conditional Expectation (ICE) Q-functions that enables joint estimation through shared representations. We implement this approach in the Policy-Encoded Q Network (PEQ-Net), an architecture centered on a shared policy encoder. The encoder is trained using kernel mean embeddings, ensuring that the learned representation space reflects population-level policy dissimilarities. After applying an LTMLE correction step, we prove this design imposes a structural constraint on the second-order remainder, thereby stabilizing finite-sample variance. Experiments on semi-synthetic datasets demonstrate that PEQ-Net consistently outperforms existing ICE-based methods, achieving substantial reductions in root-mean-square error, particularly when evaluating closely related policies.

Watermarking techniques for large language models (LLMs), which encode hidden information in the output so its source can be verified, have gained significant attention in recent days, thanks to their potential capability to detect accidental or deliberate misuse. Similar challenges involving model misuse also exist in the context of game-playing, such as when detecting the unauthorized use of AI tools in gaming platforms (e.g., cheating in online chess). In this paper, we initiate the study of how game-playing strategies can be watermarked. We show how the KGW watermark for LLMs can be adapted to watermark game-playing agents in perfect-information extensive-form games. The watermark can then be detected using a statistical test. We show that the degradation in the quality of the watermarked strategy profile, quantified by the expected utility, can be bounded, but there is a tradeoff between detectability and quality. In our experiments, we bootstrap the watermarking framework to various chess engines and demonstrate that a) the impact of the watermark on the quality of the strategy is negligible and b) the watermark can be detected with just a handful of games.

We introduce and analyze Target-Induced Loss Tilting (TILT) for unsupervised domain adaptation under covariate shift. It is based on a novel objective function that decomposes the source predictor as $f+b$, fits $f+b$ on labeled source data while simultaneously penalizing the auxiliary component $b$ on unlabeled target inputs. The resulting fit $f$ is deployed as the final target predictor. At the population level, we show that this target-side penalty implicitly induces relative importance weighting at the population level, but in terms of an estimand $b^*_f$ that is self-localized to the current error, and remains uniformly bounded for any source-target pair (even those with disjoint supports). We prove a general finite-sample oracle inequality on the excess risk, and use it to give an end-to-end guarantee for training with sparse ReLU networks. Experiments on controlled regression problems and shifted CIFAR-100 distillation show that TILT improves target-domain performance over source-only training, exact importance weighting, and relative density-ratio baselines, with a stable dependence on the regularization parameter.

Parallelization has played an instrumental role in the field of artificial intelligence (AI), drastically reducing the time taken to train and evaluate large AI models. In contrast to its impact in the broader field of AI, applying parallelization to computational game solving is relatively unexplored, despite its great potential. In this paper, we parallelize the family of counterfactual regret minimization (CFR) algorithms, which were central to important breakthroughs for solving large imperfect-information games. We present a generalized parallelization framework, reframing CFR as a series of linear algebra operations. Then, existing techniques for parallelizing linear algebra operations can be applied to accelerate CFR. We also describe how our technique can be applied to other tabular members of the CFR family of algorithms, including the state-of-the-art, such as CFR+, discounted CFR, and predictive variants of CFR. Experimentally, we show that our CFR implementation on a GPU is up to four orders of magnitude faster than Google DeepMind OpenSpiel's CFR implementations on a CPU.

Diffusion models generate samples by denoising along the score of a perturbed target distribution. In practice, one trains a neural diffusion model, which is computationally expensive. Recent work suggests that score matching implicitly smooths the empirical score, and that this smoothing bias promotes generalization by capturing low-dimensional data geometry. We propose moment-matched score-smoothed overdamped Langevin dynamics (MM-SOLD), a training-free interacting particle sampler that enforces the target moments throughout the sampling trajectory. We prove that, in the large-particle limit, the empirical particle density converges to a deterministic limit whose one-particle stationary marginal is a Gibbs--Boltzmann density obtained by exponentially tilting a naive score-smoothed diffusion target. The mean and covariance of this distribution agree with the empirical moments of the training data. Experiments on 2D distributions and latent-space image generation show that MM-SOLD enables fast, robust, training-free sampling on CPUs, with sample fidelity and diversity competitive with neural diffusion baselines.

LLM agents increasingly run inside execution harnesses that dispatch tools, allocate resources, and route messages between specialized components. However, a harness can return a correct, benign answer over a trajectory that accesses unauthorized resources or leaks context to the wrong agent. Output-level evaluation cannot see these failures, yet most safety benchmarks score only final outputs or terminal states, even though many violations occur mid-trajectory rather than at termination. The central question is whether the harness respects user intent, permission boundaries, and information-flow constraints throughout execution. To address this gap, we propose HarnessAudit, a framework that audits full execution trajectories across boundary compliance, execution fidelity, and system stability, with a focus on multi-agent harnesses where these risks are most pronounced. We further introduce HarnessAudit-Bench, a benchmark of 210 tasks across eight real-world domains, instantiated in both single-agent and multi-agent configurations with embedded safety constraints. Evaluating ten harness configurations across frontier models and three multi-agent frameworks, we find that: (i) task completion is misaligned with safe execution, and violations accumulate with trajectory length; (ii) safety risks vary across domains, task types, and agent roles; (iii) most violations concentrate in resource access and inter-agent information transfer; and (iv) multi-agent collaboration expands the safety risk surface, while harness design sets the upper bound of safe deployment.

Generating realistic human motion is a central yet unsolved challenge in video generation. While reinforcement learning (RL)-based post-training has driven recent gains in general video quality, extending it to human motion remains bottlenecked by a reward signal that cannot reliably score motion realism. Existing video rewards primarily rely on 2D perceptual signals, without explicitly modeling the 3D body state, contact, and dynamics underlying articulated human motion, and often assign high scores to videos with floating bodies or physically implausible movements. To address this, we propose PhyMotion, a structured, fine-grained motion reward that grounds recovered 3D human trajectories in a physics simulator and evaluates motion quality along multiple dimensions of physical feasibility. Concretely, we recover SMPL body meshes from generated videos, retarget them onto a humanoid in the MuJoCo physics simulator, and evaluate the resulting motion along three axes: kinematic plausibility, contact and balance consistency, and dynamic feasibility. Each component provides a continuous and interpretable signal tied to a specific aspect of motion quality, allowing the reward to capture which aspects of motion are physically correct or violated. Experiments show that PhyMotion achieves stronger correlation with human judgments than existing reward formulations. These gains carry over to RL-based post-training, where optimizing PhyMotion leads to larger and more consistent improvements than optimizing existing rewards, improving motion realism across both autoregressive and bidirectional video generators under both automatic metrics and blind human evaluation (+68 Elo gain). Ablations show that the three axes provide complementary supervision signals, while the reward preserves overall video generation quality with only modest training overhead.

Diffusion models (DMs) have exhibited remarkable efficacy in various image restoration tasks. However, existing approaches typically operate within the high-dimensional pixel space, resulting in high computational overhead. While methods based on latent DMs seek to alleviate this issue by utilizing the compressed latent space of a variational autoencoder, they require repeated encoder-decoder inference. This introduces significant additional computational burdens, often resulting in runtime performance that is even inferior to that of their pixel-space counterparts. To mitigate the computational inefficiency, this work proposes projecting data into lower-dimensional subspaces using dynamic resolution DMs to accelerate the inference process. We first fine-tune pre-trained DMs for dynamic resolution priors and adapt DPS and DAPS, which are two widely used pixel-space methods for general image restoration tasks, into the proposed framework, yielding methods we refer to as SubDPS and SubDAPS, respectively. Given the favorable inference speed and reconstruction fidelity of SubDAPS, we introduce an enhanced variant termed SubDAPS++ to further boost both reconstruction efficiency and quality. Empirical evaluations across diverse image datasets and various restoration tasks demonstrate that the proposed methods outperform recent DM-based approaches in the majority of experimental scenarios. The code is available at https://github.com/StarNextDay/SubDAPS.git.

Integration of artificial intelligent (AI) agents in higher education is transforming teaching, learning and administrative processes. Although existing AI agents effectively support individual tasks, their implementation remains fragmented and inefficient for handling the complexity of educational institutions. This highlights a significant research gap: the lack of integrated eco-system-level agentic multi-agent AI platform capable of coordinated planning, reasoning, and adaptive decision-making across multiple educational functions. This paper presents a forward-looking perspective on agentic multi-agent AI platform in higher education, consisting interconnected autonomous, goal driven agents that support learning, teaching, and institutional operations. It addresses timely and critical questions: Can agentic AI represent the next generation of intelligent systems in tertiary education? Can they collectively support seamless coordinated operations across teaching, learning and administrative support? To what extent can such systems foster inclusive and equitable learning for diverse learners with special educational needs? To ground this perspective, a thematic analysis of existing literature identifies four dominant themes: task-specific fragmented AI tools, the transition from single-agent to multi-agent systems, limited cross-functional integration, and insufficient focus on inclusivity and accessibility. Findings reveal a clear gap between current AI implementations and the needs of holistic, learner-centered educational ecosystem. The paper synthesizes challenges and outlines future research directions for scalable human-aligned, and inclusive agentic AI platform. The significant contribution is the incorporation of inclusive learning perspectives, highlighting how coordinated agentic multi-agent platform can support diverse learners through adaptive, multimodal interventions.

How should an agent's performance in a multiagent environment be evaluated when there is a limited sample size or a high cost of running a trial? The AIVAT family of variance reduction techniques was proposed to address this challenge by introducing unbiased low-variance estimators of agents' expected payoffs. An important component of AIVAT is a heuristic value function that discriminates between potentially low- and high-value counterfactual histories. A notable gap in the literature is that there is little to no constraint or guideline on how the heuristic value function should be chosen or how uncertainty in its output should be handled. In our first contribution, we parameterize the heuristic value function to highlight AIVAT's potential vulnerabilities: a) the sample variance can be set pathologically low by directly applying gradient descent on the sample variance, and b) one can p-hack to draw a desired statistical conclusion via gradient descent/ascent on the test statistic. The main takeaway is that the heuristic value function should be fixed prior to observing the evaluation data! In our second contribution, we show how the heuristic uncertainty can be propagated to quantify the uncertainty of AIVAT estimates. It is then possible to further reduce the variance using inverse-variance weighted averaging, but AIVAT's unbiasedness guarantee may have to be sacrificed. In our experiments, we use a dataset of 10,000 poker hands to demonstrate our heuristic pathology and uncertainty results, with the latter yielding a 43.0% reduction in the number of samples (poker hands) needed to draw statistical conclusions.

Conformal prediction is often calibrated with a single pooled threshold, but this can hide cross-group heterogeneity in score distributions and distort group-wise coverage. We study this phenomenon through the population score distributions underlying split conformal calibration. First, we derive a conservation law and lower bound showing that pooled calibration incurs irreducible group-wise coverage distortion at a scale set by cross-group quantile heterogeneity. Second, we demonstrate that the two leading fairness definitions for conformal prediction, Equalized Coverage and Equalized Set Size, are fundamentally in tension. Third, we quantify the cost of moving between policies which treat groups separately or pool them. Experiments on synthetic and real data confirm the same bidirectional trade-off after finite-sample calibration. Our results show that, for the policy families studied here, calibration choice does not remove cross-group heterogeneity; it determines whether the resulting distortion appears in the coverage or size dimension, providing a principled lens for analyzing fairness-oriented calibration choices in practice.

Applying Large Language Models (LLMs) to heterogeneous enterprise systems is hindered by hallucinations and failures in multi-hop, n-ary reasoning. Existing paradigms (e.g., GraphRAG, NL2SQL) lack the semantic grounding and auditable execution required for these complex environments. We introduce HEAR, an enterprise agentic reasoner built on a Stratified Hypergraph Ontology. Its base Graph Layer virtualizes provenance-aware data interfaces, while the Hyperedge Layer encodes n-ary business rules and procedural protocols. Operating an evidence-driven reasoning loop, HEAR dynamically orchestrates ontology tools for structured multi-hop analysis without requiring LLM retraining. Evaluations on supply-chain tasks, including order fulfillment blockage root cause analysis (RCA), show HEAR achieves up to 94.7% accuracy. Crucially, HEAR demonstrates adaptive efficiency: utilizing procedural hyperedges to minimize token costs, while leveraging topological exploration for rigorous correctness on complex queries. By matching proprietary model performance with open-weight backbones and automating manual diagnostics, HEAR establishes a scalable, auditable foundation for enterprise intelligence.

Large language models are remarkably capable, yet how computation propagates through their layers remains poorly understood. A growing line of work treats depth as discrete time and the residual stream as a dynamical system, where each layer's nonlinear update has a local linear description. However, previous analyses have relied on scalar summaries or approximate linearizations, leaving the full spectral geometry of trained LLMs unknown. We perform full Jacobian eigendecomposition across three production--scale LLMs and show that training installs a monotonic spectral gradient through depth -- from non-normal, rotation-dominated early layers to near--symmetric late layers -- together with a cumulative low-rank bottleneck that funnels perturbations into a small fraction of the residual stream's effective dimensions. Our experiments reveal that this gradient and the dimensional collapse are learned rather than architectural, and is largely dissolved when structured non-normality is removed. We further show that the topological positioning of graph communities predicts whether the Jacobian amplifies or suppresses them, with the sign of the coupling determined by the local operator type, a relationship absent at initialization. These results map a learned spectral geometry in LLMs that links perturbation propagation and compression to the network's functional topology.

We describe two types of models for vocabulary difficulty prediction: a high-accuracy black-box model, which achieved the top shared task result in the open track, and an explainable model, which outperforms a fine-tuned encoder baseline. As the black-box model, we fine-tuned an LLM using a soft-target loss function for effective application to the rating task, achieving r > 0.91. The explainable model provides insights into what impacts the difficulty of each item while maintaining a strong correlation (r > 0.77). We further analyze the results, demonstrating that the difficulty of items in the British Council's Knowledge-based Vocabulary Lists (KVL) is often affected by spelling difficulty or the construction of the test items, in addition to the genuine production difficulty of the words. We make our code available online at https://github.com/adno/vocabulary-difficulty .

Industrial visual inspection systems increasingly rely on deep classifiers whose heatmap explanations may appear visually plausible while failing to identify the image regions that actually drive model decisions. This paper operationalizes an architecture-aware explanation audit protocol grounded in the native-readout hypothesis: the perturbation-based faithfulness of an explanation method is bounded by its structural distance from the model's native decision mechanism. On WM-811K wafer maps (9 classes, 172k images) under a three-seed zero-fill perturbation protocol, ViT-Tiny + Attention Rollout attains Deletion AUC 0.211 against 0.432-0.525 for Swin-Tiny / ResNet18+CBAM / DenseNet121 + Grad-CAM (abs(Cohen's d) > 1.1), despite lower classification accuracy. Swin-Tiny disentangles architecture family from readout structure: despite being a Transformer, its spatial feature-map hierarchy makes it Grad-CAM compatible, showing that the operative factor is readout structure rather than architecture family. A model-agnostic control (RISE) compresses all families to Deletion AUC about 0.1, indicating the gap arises from the explainer pathway; notably, RISE outperforms all native methods, so native readout is a compatibility principle rather than an optimality guarantee. A blur-fill sensitivity analysis shows that the family ordering reverses under a different perturbation baseline, reinforcing that faithfulness rankings are joint properties of (model, explainer, perturbation operator) triples. An exploratory boundary-condition study on MVTec AD (pretrained models) indicates that audit results are dataset/task dependent and identifies conditions requiring qualification. The protocol yields actionable guidance: explanation pathways should be co-designed with model architectures based on readout structure, and deployed heatmaps should be accompanied by quantitative faithfulness metrics.

Purpose: Mechanical thrombectomy (MT) improves stroke outcomes, but is limited by a lack of local treatment access. Widespread distribution of reinforcement learning (RL)-based robotic systems can be used to alleviate this challenge through autonomous navigation, but current RL methods require live device tip coordinate tracking to function. This paper aims to develop and evaluate a real-time catheter tip tracking pipeline under fluoroscopy, addressing challenges such as low contrast, noise, and device occlusion. Methods: A multi-threaded pipeline was designed, incorporating frame reading, preprocessing, inference, and post-processing. Deep learning segmentation models, including U-Net, U-Net+Transformer, and SegFormer, were trained and benchmarked using two-class and three-class formulations. Post-processing involved two-step component filtering, one-pixel medial skeletonization, and greedy arc-length path following with contour fall-back. Results: On manually-labeled moderate complexity fluoroscopic video data, the two-class SegFormer achieved a mean absolute error of 4.44 mm, outperforming U-Net (4.60 mm), U-Net+Transformer (6.20 mm) and all three-class models (5.19-7.74 mm). On segmentation benchmarks, the system exceeded state-of-the-art CathAction results with improvements of up to +5% in Dice scores for three-segmentation. Conclusion: The results demonstrate that the proposed multi-threaded tracking framework maintains stable performance under challenging imaging conditions, outperforming prior benchmarks, while providing a reliable and efficient foundation for RL-based autonomous MT navigation.

Spiking neural networks (SNNs), which are brain-inspired and spike-driven, achieve high energy efficiency. However, a performance gap between SNNs and artificial neural networks (ANNs) still remains. Knowledge distillation (KD) is commonly adopted to improve SNN performance, but existing methods typically enforce uniform alignment across all timesteps, either from a teacher network or through inter-temporal self-distillation, implicitly assuming that per-timestep predictions should be treated equally. In practice, SNN predictions vary and evolve over time, and intermediate timesteps need not all be individually correct even when the final aggregated output is correct. Under such conditions, effective distillation should not force every timestep toward the same supervision target, but instead provide corrective guidance to erroneous timesteps while preserving useful temporal dynamics. To address this issue, we propose Selective Alignment Knowledge Distillation (SeAl-KD), which selectively aligns class-level and temporal knowledge by equalizing competing logits at erroneous timesteps and reweighting temporal alignment based on confidence and inter-timestep similarity. Extensive experiments on static image and neuromorphic event-based datasets demonstrate consistent improvements over existing distillation methods. The code is available at https://github.com/KaiSUN1/SeAl

We present EnergyLens, an end-to-end framework for energy-aware large language model (LLM) inference optimization. As LLMs scale, predicting and reducing their energy footprint has become critical for sustainability and datacenter operations, yet existing approaches either require production-level code and expensive profiling or fail to accurately capture multi-GPU energy behavior. As a result, practitioners lack tools for deciding which optimizations to prioritize and for selecting among existing deployment configurations when exhaustive profiling is impractical. EnergyLens addresses this gap with an intuitive einsum-based interface that captures LLM specifications including fusion, parallelism, and compute-communication overlap, combined with load-imbalance-aware MoE modeling and an empirically driven communication energy model for multi-GPU settings. We validate EnergyLens on Llama3 and Qwen3-MoE across tensor-parallel and expert-parallel configurations, achieving mean absolute percentage errors (MAPEs) between 9.25% and 13.19% for multi-GPU prefill and decode energy, and 12.97% across SM allocations for Megatron-style overlap. Our energy-driven exploration reveals up to 1.47x and 52.9x energy variation across configurations in prefill and decode efficiency and motivates distributed serving. We further show that compute-communication overlap is difficult to optimize with intuition alone, but EnergyLens correctly identifies Pareto-optimal overlap configurations.

Many safety-critical control problems are modeled as risk-sensitive partially observable Markov decision processes, where the controller must make decisions from incomplete observations while balancing task performance against safety risk. Although belief-space planning provides a principled solution, maintaining and planning over beliefs can be computationally costly and sensitive to model specification in practical domains. We propose a lightweight risk-gated reinforcement learning approximation for risk-sensitive control under partial observability. The method constructs a compact finite-history proxy state and learns an action-conditioned predictor of near-term safety violation. This predicted candidate-action risk is used in two complementary ways: as a risk penalty during value learning, and as a decision-time gate that interpolates between optimistic and conservative ensemble value estimates. As a result, low-risk actions are evaluated closer to reward-seeking estimates, while high-risk actions are evaluated more conservatively. We evaluate the approach in two safety-critical partially observable domains: automated glucose regulation and safety-constrained navigation. Across adult and adolescent glucose-control cohorts, the method improves overall glycemic tradeoffs and substantially reduces runtime relative to a belief-space planning baseline. On Safety-Gym navigation benchmarks, it achieves a more favorable reward-cost balance than unconstrained RL and several standard safe-RL baselines. These results suggest that action-conditioned near-term risk can provide an effective local signal for approximate risk-sensitive POMDP control when full belief-space planning is impractical.

The monitoring of fetal heart rate (FHR) and the assessment of its variability are crucial for preventing fetal compromise and adverse outcomes. However, traditional methods encounter limitations arising from equipment performance, data transmission, and subjective assessments by doctors. We have developed a tailored AI-based FHrCTG model specifically for FHR monitoring, which effectively mitigates noise interference and precisely reconstructs signals. Our model was pre-trained on a massive dataset consisting of 558,412 unlabeled data points and further refined using 7,266 expert-reviewed entries. To validate FHR, we introduced the Intersection Overlapping Labels (IOL) approach, which transforms rate analysis into categorical judgments. Testing revealed that our model demonstrates high sensitivity and specificity in detecting critical FHR decelerations (89.13% and 87.78%, respectively) and accelerations (62.5% and 92.04%, respectively). Furthermore, based on Fischer's criteria for clinical application, our model achieved impressive AUC scores of 0.7214 and 0.9643 for verifying FHR periodicity and amplitude variation, respectively.

Tool-augmented LLM agents increasingly access the same tool type through multiple functionally equivalent providers, such as web-search APIs, retrievers, or LLM backends exposed behind a shared interface. This creates a provider-routing problem under runtime load: the router must choose among providers that differ in latency, reliability, and answer quality, often without gold labels at deployment time. We introduce LQM-ContextRoute, a contextual bandit router for same-function tool providers. Its key design is latency-quality matching: instead of letting low latency offset poor answers in an additive reward, the router ranks providers by expected answer quality per service cycle. It combines this capacity-aware score with query-specific quality estimation and LLM-as-judge feedback, allowing it to adapt online to both load changes and provider-quality differences. On the main web-search load benchmark, LQM-ContextRoute improves F1 by +2.18 pp over SW-UCB while staying on the latency-quality frontier. In a high-heterogeneity StrategyQA setting, LQM-ContextRoute avoids additive-reward collapse and improves accuracy by up to +18 pp over SW-UCB; on heterogeneous retriever pools, it improves NDCG by +2.91--+3.22 pp over SW-UCB. These results show that same-function tool routing benefits from treating latency as service capacity, especially when runtime pressure and provider-quality heterogeneity coexist.

The recent large-scale emergence of LLMs has left an open space for dealing with their consequences, such as plagiarism or the spread of false information on the Internet. Coupling this with the rise of AI detector bypassing tools, reliable machine-generated text detection is in increasingly high demand. We investigate the paraphrasing attack resilience of various machine-generated text detection methods, evaluating three approaches: fine-tuned RoBERTa, Binoculars, and text feature analysis, along with their ensembles using Random Forest classifiers. We discovered that Binoculars-inclusive ensembles yield the strongest results, but they also suffer the most significant losses during attacks. In this paper, we present the dichotomy of performance versus resilience in the world of AI text detection, which complicates the current perception of reliability among state-of-the-art techniques.

Deploying AI agents for repetitive periodic tasks exposes a critical tension: Large Language Models (LLMs) offer unmatched flexibility in tool orchestration, yet their inherent stochasticity causes unpredictable failures, and repeated invocations incur prohibitive token costs. We present the LOOP SKILL ENGINE, a system that achieves a combined 99% success rate and 99% token reduction for periodic agent tasks through a one-shot recording, deterministic replay paradigm. On its first run, the agent executes the task with full LLM reasoning while the system transparently intercepts and records the complete tool-call trajectory. A greedy length-descending template extraction algorithm then converts this recording into a parameterized, branch-free Loop Skill -- a deterministic execution plan that captures the task's functional intent while parameterizing time-dependent and result-dependent variables. All subsequent executions bypass the LLM entirely: the engine resolves template variables against real-time values and replays the tool sequence deterministically. We prove two theorems: (1) Replay Determinism -- the step sequence of a validated Loop Skill is invariant across all future executions; (2) Write Safety -- concurrent access to persistent configuration is serialized through reentrant locks and atomic file replacement. Across a benchmark of periodic agent tasks spanning intervals from 5 minutes to 24 hours, the Loop Skill Engine reduces monthly token consumption by 93.3%--99.98% and cuts execution latency by 8.7x while eliminating output non-determinism. A multi-layer degradation strategy guarantees that tasks never stall. We release the engine as part of the buddyMe open-source agent framework.

Pairwise Ranking Prompting (PRP) elicits pairwise preference judgments from an LLM, which are then aggregated into a ranking, usually via classical sorting algorithms. However, judgments are noisy, order-sensitive, and sometimes intransitive, so sorting assumptions do not match the setting. Because sorting aims to recover a full permutation, truncating it to meet a call budget does not produce a dependable top-K. We thus reframe PRP reranking as active learning from noisy pairwise comparisons and show that active rankers are drop-in replacements that improve NDCG@10 per call in the call-constrained regime. Our noise-robust framework also introduces a randomized-direction oracle that uses a single LLM call per pair. This approach converts systematic position bias into zero-mean noise, enabling unbiased aggregate ranking without the cost of bidirectional calls.

We present an empirical evaluation of quantum entanglement in agent coordination within quantum multi agent reinforcement learning (QMARL). While QMARL has attracted growing interest recently, most prior work evaluates quantum policies without provable baselines, making it impossible to rigorously distinguish quantum advantage from algorithmic coincidence. We address this directly by evaluating a decentralized QMARL framework with variational quantum circuit (VQC) actors with shared entangled states. In the CHSH game, which has a mathematically proven classical performance ceiling of 0.75 win rate, we show that entangled QMARL agents approach the Tsirelson limit of 0.854, providing clear evidence of their quantum advantage. We show that unentangled quantum circuits match the classical baseline, confirming that entanglement and not the quantum circuit itself is the active coordination mechanism. We also explore the effect of specific entanglement structures, as some Bell states enable coordination gains while others actively harm performance. On cooperative navigation (CoopNav), QMARL without entanglement achieves $\sim2\times$ improvement in success rate over classical MAA2C ($\sim$0.85 versus $\sim$0.40), with the hybrid configuration, quantum actor paired with a classical centralised critic, outperforming both fully classical and fully quantum solutions. We present our experimental analysis and discuss future work.

Audio self-supervised learning (SSL) aims to learn general-purpose representations from large-scale unlabeled audio data. While recent advances have been driven mainly by generative reconstruction objectives, contrastive approaches remain less explored, partly due to the difficulty of designing effective audio augmentations and the large batch sizes required for contrastive pre-training. We introduce \textbf{AudioMosaic}, a contrastive learning-based audio encoder for general audio understanding. During pre-training, AudioMosaic constructs positive pairs by applying structured time-frequency masking to spectrogram patches, which reduces memory usage and enables efficient large-batch training. Compared with generative approaches, the AudioMosaic encoder learns more discriminative utterance-level representations that demonstrate strong transferability across datasets, domains, and acoustic conditions. Extensive experiments show that AudioMosaic achieves state-of-the-art performance on several standard audio benchmarks under both linear probing and fine-tuning. We further show that integrating the pretrained AudioMosaic encoder into audio-language models improves performance on audio-language tasks. The code is publicly available in our \href{https://github.com/HanxunH/AudioMosaic}{GitHub repository}.

Background: Abilities for effective self-regulated learning (SRL) are critical for lifelong learning, particularly during adolescence when these skills consolidate and strongly influence future learning. Their importance has grown with the rise of online and blended education. Yet, little is known about how secondary school students self-regulate in online environments, how their SRL processes and strategies evolve, or how they affect outcomes. In secondary education, understanding these processes can reveal patterns and indicators of learning success, informing the design of online support mechanisms. Evidence from repeated-measures designs remains scarce. Objectives: This study aims to examine how secondary school students enact SRL strategies during online essay writing, how these strategies change over time, and how they relate to learning outcomes. Methods: We analysed metacognition-related trace data collected from secondary students during a two-wave online essay-writing task conducted one week apart in two Colombian schools (N = 93 for session 1, N = 95 for session 2) via a digital learning platform. Using a combination of process mining and unsupervised machine learning techniques, we identified dominant SRL strategies grounded in established SRL processes and examined their stability and association with learning outcomes. Results and conclusions: Three dominant SRL strategies were identified. Results showed variability: many students remained in or shifted to Read first, write next, while none used Write intensively, read selectively in session 2. Although less common, latter strategy was positively associated with learning outcomes.

Accurate disease trajectory prediction is critical for early intervention, resource allocation, and improving long-term outcomes. While electronic health records (EHRs) provide a rich longitudinal view of patient health in clinical environments, models trained on curated research cohorts may not reflect routine deployment settings, and those trained on single-hospital datasets capture only fragments of each patient's trajectory. This highlights the importance of leveraging large, multi-hospital health systems for training and validation to better reflect real-world clinical complexity. In this work, we develop DT-Transformer, a foundation model trained on 57.1M structured EHR entries over 1.7M patients from Mass General Brigham (MGB), spanning 11 hospitals and a broad network of outpatient clinics. DT-Transformer achieves strong discrimination in both held-out and prospective validation settings. Next-event prediction achieves a median age- and sex-stratified AUC of 0.871 across 896 disease categories, with all categories exceeding AUC 0.5. These results support health system-scale training as a path toward foundation models suited to real-world clinical forecasting.

We present an in-depth analysis of the Koopman semigroup via wavelet transform. Towards this goal, we start by introducing the wavelet-based observables and show that they are eigenfunctions of the Koopman semigroup when this semigroup is considered over the Banach space of continuous functions on a compact forward-invariant set endowed with the supremum norm. We then construct closed-form expressions of the action of the Koopman semigroup and its resolvent in terms of these observables. To approximate the action of Koopman semigroup numerically, we combine Extended Dynamic Mode Decomposition (EDMD) with the proposed wavelet-based observables leading to the Wavelet Dynamic Mode Decomposition via Continuous Wavelet Transform (cWDMD) algorithm. We validate our theoretical results on two numerical examples.

Modern LLM RL systems separate rollout generation from policy optimization. These two stages are expected to produce token probabilities that match exactly. However, implementation differences can make them assign different values to the same sequence under the same model weights, inducing Training-Inference Mismatch (TIM). TIM is difficult to inspect because it is entangled with off-policy drift and common stabilization mechanisms. In this work, we isolate TIM in a zero-mismatch diagnostic setting (VeXact), and show that small token-level numerical disagreements can independently cause training collapse. We further show that TIM changes the effective optimization problem, and identify a set of remedies that could mitigate TIM. Our results suggest that TIM is not benign numerical noise, but a systems-level perturbation that should be treated as a first-order factor in analyzing LLM RL stability.

Hybrid quantum-classical (HQC) algorithms, such as the Variational Quantum Eigensolver (VQE) and the Quantum Approximate Optimization Algorithm (QAOA), are central to near-term quantum computing but remain challenging to test. Sampling-based fuzzing can expose faulty or non-convergent configurations, but under realistic execution budgets, it may miss failure-prone regions in the joint space of classical optimizer settings and quantum circuit parameters. This paper studies failure-guided fuzzing for HQC programs. It models a hybrid input as a pair of classical optimizer hyperparameters and quantum circuit parameters, and evaluates a two-phase strategy that first searches for non-convergent seeds and then locally fuzzes circuit parameters around those seeds. To understand where the gains come from, five budgeted strategies are compared: random hybrid testing, classical enumeration without fuzzing, random-seed local fuzzing, enumeration-seed local fuzzing, and concolic-seed local fuzzing. The study is implemented on a VQE instance and a QAOA MaxCut instance in Qiskit. The results show that failure-guided local fuzzing is the main driver of improvement over random testing, while concolic seed discovery provides additional benefits on VQE but is less stable on QAOA. These findings suggest that reusing failure information is a promising direction for HQC testing, but that the value of concolic seed discovery is workload-dependent.

The key problem facing ChatGPT-like AI's use across society is that its behavior can shift, unnoticed, from desirable to undesirable -- encouraging self-harm, extremist acts, financial losses, or costly medical and military mistakes -- and no one can yet predict when. Shifts persist in even the newest AI models despite remarkable progress in AI modeling, post-training alignment and safeguards. Here we show that a vector generalization of fusion-fission group dynamics observed in living and active-matter systems drives -- and can forecast -- future shifts in the AI's behavior. The shift condition, which is also derivable mathematically, results from group-level competition between the conversation-so-far (C) and the desirable (B) and undesirable (D) basin dynamics which can be estimated in advance for a given application. It is neither model-specific nor driven by stochastic sampling. We validate it across six independent tests, including: 90 percent correct across seven AI models spanning two orders of magnitude in parameter count (124M-12B); production-scale persistence across ten frontier chatbots; and a priori time-stamped prediction eleven months before the Stanford 'Delusional Spirals' corpus appeared, and independently confirmed by that corpus of 207,443 human-AI exchanges. Because it sits architecturally below the current safety stack, the same formula provides a real-time warning signal that current alignment does not supply, portable across current and future ChatGPT-like AI architectures and instantiable in application domains where competing response classes can be defined.

Large language models can now be personalised efficiently at scale using parameter efficient finetuning methods (PEFTs), but serving user-specific PEFTs harms throughput, even with specialised kernels and memory management techniques. This is because, theoretically and empirically, a mismatch exists between prefill (processing a large number of tokens at once) and decode (generating a single token autoregressively): the latter has far lower throughput when serving multiple adapters. Rather than optimising performance relative to parameter count, for efficient multi-adapter serving, we instead ought to optimise performance relative to serving throughput. We therefore propose PreFT (Prefill-only Finetuning), wherein we only apply the adapter to prefill tokens and discard it afterwards. PreFT significantly increases throughput with minimal effect on performance. We develop and release an efficient implementation of two prefill-only PEFTs, LoRA and ReFT, on the vLLM inference engine. We first show that serving multi-user PreFTs is more efficient than traditional PEFTs ($1.9\times$ the throughput when serving $512$ adapters on Llama 3.1 70B). Then, we compare the performance of prefill-only vs. all-token adapters on a variety of supervised finetuning and reinforcement learning tasks with LMs at varying scales. On SFT, we observe that the evaluation loss of PreFTs is higher than PEFTs, but can be compensated by increasing rank with nearly no reduction in throughput. On RL, we consistently find that PreFTs approach parity with standard PEFTs. Together, this work validates prefill-only adaptation of LLMs as a more favourable accuracy-throughput tradeoff than existing PEFTs for personalised serving.

Genetic circuit design remains a laborious, expert-driven process despite decades of progress in synthetic biology. We study this problem through code generation: models produce Python code in pysbol3 to construct genetic circuits in the Synthetic Biology Open Language (SBOL), a formal representation that supports automated verification. We introduce GenCircuit-RL, a reinforcement learning framework built around hierarchical verification rewards that decompose correctness into five levels, from code execution to task-specific topological checks, and a four-stage curriculum that shifts optimization pressure from code generation to functional reasoning. We also introduce SynBio-Reason, a benchmark of 4,753 circuits spanning six canonical circuit types and nine tasks from code repair to de novo design, with held-out biological parts for out-of-distribution evaluation. Hierarchical verification improves task success on functional reasoning tasks by 14 to 16 percentage points over binary rewards, and curriculum learning is required for strong design performance. The resulting models generate topologically correct circuits, generalize to novel biological parts, and rediscover canonical designs from the synthetic biology literature.

Automatic multi-agent systems aim to instantiate agent workflows without relying on manually designed or fixed orchestration. However, existing automatic MAS approaches remain only partially adaptive: they either perform training-free test-time search or optimize the meta-level designer while keeping downstream execution agents frozen, which creating a frozen-executor ceiling and leaving the end-to-end training of self-designing and self-executing agentic models unexplored. To address this, we introduce MetaAgent-X, an end-to-end reinforcement learning framework that jointly optimizes automatic MAS design and execution. MetaAgent-X enables script-based MAS generation, execution rollout collection, and credit assignment for both designer and executor trajectories. To support stable and scalable optimization, we propose Executor Designer Hierarchical Rollout and Stagewise Co-evolution to improve training stability and expose the dynamics of designer-executor co-evolution. MetaAgent-X consistently outperforms existing automatic MAS baselines, achieving up to 21.7% gains. Comprehensive ablations show that both designer and executor improve throughout training, and that effective automatic MAS learning follows a stagewise co-evolution process. These results establish end-to-end trainable automatic MAS as a practical paradigm for building self-designing and self-executing agentic models.

Long-horizon embodied tasks remain a fundamental challenge in AI, as current methods rely on hand-engineered rewards or action-labeled demonstrations, neither of which scales. We introduce ASH, an agentic system that learns an embodied policy from unlabeled, noisy internet video, without reward shaping or expert annotation. ASH follows a self-improvement loop; when it gets stuck, ASH learns an Inverse Dynamics Model (IDM) from its own trajectories, and uses its IDM to extract supervision from relevant internet video. ASH uses unsupervised learning to identify key moments from large-scale internet video and retains them as long-term memory -- allowing it to tackle long-horizon problems. We evaluate ASH on two complementary environments demanding multi-hour planning: Pokemon Emerald, a turn-based RPG, and The Legend of Zelda: The Minish Cap, a real-time action-adventure game. In both games, behavioral cloning, retrieval-augmented and zero-shot foundation-model baselines plateau, while ASH sustains progression across our 8-hour evaluation. ASH reaches an average of $11.2/12$ milestones in Pokemon Emerald and $9.9/12$ in Legend of Zelda, while the strongest baseline gets stuck in both environments at an average of $6.5/12$ and $6.0/12$ milestones, respectively. We demonstrate that self-improving agents are a scalable recipe for long-horizon embodied learning.

Concept Bottleneck Models (CBMs) offer interpretable alternatives to black-box predictors by introducing human-relatable concepts before the final output. However, existing CBMs struggle to verify whether predicted concepts correspond to the correct visual evidence, limiting their reliability. We propose a fine-grained CBM framework that grounds each concept in localized visual evidence, enabling direct inspection of where and how concepts are encoded. This design allows users to interpret predictions and verify that the model learns intended concepts rather than spurious correlations. Experiments on medical imaging benchmarks show that our learned concept space is information-complete and achieves predictive performance comparable to standard CBMs, while substantially improving transparency. Unlike post-hoc attribution methods, our framework validates both the presence and correctness of concept representations, bridging interpretability with verifiability. Our approach enhances the trustworthiness of CBMs and establishes a principled mechanism for human-model interaction at the concept level, paving the way toward more reliable and clinically actionable concept-based learning systems.

LLM-based web agents can navigate live storefronts, yet they often collapse to a single "average buyer" policy, failing to capture the heterogeneous and distributional nature of real buyer populations. Existing personalization methods rely on hand-crafted prompt-based personas that are brittle, difficult to scale, context-inefficient, and unable to faithfully represent population-level behavior. We introduce SimPersona, a novel framework that learns discrete buyer types from historical traffic and exposes them to LLM-based web agents as compact persona tokens. Given raw clickstreams, a behavior-aware VQ-VAE induces a discrete buyer-type space that captures the statistical structure of real buyer behavior and merchant-specific buyer population distributions. To provide behavior-specific guidance to LLM-based web agents, SimPersona maps each learned buyer type to a dedicated persona token in the LLM agent vocabulary and fine-tunes the agent with these tokens on real browsing traces. At inference, each synthetic buyer is assigned to a learned buyer type with a single encoder forward pass, requiring no retraining or store-specific prompt engineering. For population-level simulation, SimPersona samples buyer types from each merchant's empirical distribution over the learned VQ-VAE codebook and instantiates agents with the corresponding persona tokens, preserving merchant-specific buyer population distributions. Evaluated on $8.37$M buyers across $42$ held-out live storefronts, SimPersona achieves $78\%$ conversion-rate alignment with real buyers, exhibits interpretable behavioral variation across buyer types, and outperforms a baseline with $8\times$ more parameters on goal-oriented shopping tasks. We further release an open-source data pipeline that converts raw e-commerce event logs into buyer representations and agent-training traces.

Malformed, missing, or boundary-value inputs in microservice APIs can cascade across dependent services, threatening reliability. Robustness testing systematically exercises such inputs to expose server-side failures, but generating diverse, effective tests remains challenging. Large Language Models can generate such tests from API specifications; however, it is unknown whether different models and prompt strategies produce diverse failure sets or converge on the same failures. We report a controlled experiment applying 7 prompt strategies to 3 open-source LLMs (14B-70B parameters) targeting 2 architecturally distinct microservice systems: one Java monolingual (6 services, 9 failure modes) and one polyglot (27 services, 14 failure modes), yielding 38 valid runs and 663 generated tests. We find that prompt strategy explains more variation in diversity than model size: a Structured prompt collapses diversity entirely, while a single model varied across three prompt strategies achieves complete failure-mode coverage on one system, outperforming any multi-model ensemble under a fixed prompt. We introduce two strategies, Guided and GuidedFewShot, that embed a mutation taxonomy from prior robustness testing research as domain context. GuidedFewShot achieves the highest single-run coverage on both systems (5 of 9 and 8 of 14 failure modes) while maintaining low cross-model similarity. A key lesson is that taxonomy rules alone are insufficient: LLMs cannot distinguish key-absent from value-empty mutations without concrete examples. Findings replicate across both systems.

Recent frontier large language models predominantly rely on Mixture-of-Experts (MoE) architectures. Despite empirical progress, there is still no principled understanding of how hyperparameters should scale with network width $N$, expert width $N_e$, number of experts $M$, sparsity $K$, and depth $L$ to ensure both stability and optimal performance at scale. We take a principled step toward resolving this gap by analyzing three different scaling regimes: (I) co-scaling $N\asymp N_e$, (II) co-scaling $N\asymp M\asymp K$, and (III) full proportional scaling of $N, N_e, M$, and $K$. For each regime, we develop a novel Dynamical Mean Field Theory (DMFT) description of the limiting training dynamics of MoEs that provides a formal foundation for our analysis. Within this framework, we derive the unique parameterization for SGD and Adam satisfying all maximal-update ($μ$) desiderata. We then show that the resulting $μ$P prescription does not reliably induce monotonic improvement with scale or robust learning-rate transfer. We trace these pathologies to scale-dependent observables in the aggregation dynamics, which motivates a refined set of desiderata that we term maximal scale stability. Guided by this principle, we derive a Maximally Scale-Stable Parameterization (MSSP) for both SGD and Adam in all three scaling regimes, and characterize the corresponding limiting dynamics - qualitatively distinct from the $μ$P limit - through a separate DMFT analysis. Experiments verify that MSSP robustly recovers learning rate transfer and monotonic improvement with scale across regimes. Combined with existing depth-scaling theory, these results provide a complete scaling prescription for MoE architectures as a function of width, depth, expert width, and number of experts.

The classic online stochastic matching problem typically requires immediate and irrevocable matching decisions. However, in many modern decentralized systems such as real-time ride-hailing and distributed cloud computing, the primary bottleneck is often local communication bandwidth rather than the timing of the match itself. We formalize this challenge by introducing a two-stage local sparsification framework. In this setting, arriving requests must prune their realized compatibility sets to a strict budget of $k$ edges before a central coordinator optimizes the global matching. This creates a "middle ground" between local information constraints and global optimization utility. We propose a local selection strategy, parametrized by a fractional solution of the expected instance. Theoretically, we quantify the approximation ratio as a function of the solution's {\em spread}. We prove that under sufficient spread, our sparsifier globally preserves the expected size of the maximum matching. Empirically, we demonstrate the robustness of our approach using the New York City ride-hailing datasets and adversarial synthetic benchmarks. Our results show that near-optimal global matching is achievable even with highly constrained local budgets, significantly outperforming standard online baselines.

Large Language Models (LLMs) pose a significant risk of safety misalignment after finetuning, as models can be compromised by both explicitly and implicitly harmful data. Even some seemingly benign data can inadvertently steer a model towards misaligned behaviors. To address this, we introduce GradShield, a principled filtering method that safeguards LLMs during finetuning by identifying and removing harmful data points before they corrupt the model's alignment. It removes potentially harmful data by computing a Finetuning Implicit Harmfulness Score (FIHS) for each data point and employs an adaptive thresholding algorithm. We apply GradShield to multiple utility fine-tuning tasks across varying levels of harmful data and evaluate the safety and utility performance of the resulting LLMs using various metrics. The results show that GradShield outperforms all baseline methods, consistently maintaining an Attack Success Rate (ASR) below $6\%$ while preserving utility performance.

Retrieval-Augmented Generation (RAG) has become a powerful and widely used approach for improving large language models by grounding generation in retrieved evidence. However, RAG systems still produce incorrect answers in many cases. Why RAG fails despite having access to external information remains poorly understood. We present a model-internal study of retrieval-augmented generation that examines how retrieved evidence influences answer generation. Using circuit tracing, we construct attribution graphs that model the flow of information through transformer layers during decoding. These graphs represent interactions among retrieved context, intermediate model activations, and generated tokens, providing a graph, circuit-level view of how external evidence is integrated into the model's reasoning process across multiple question answering benchmarks, we observe consistent structural differences: correct predictions exhibit deeper reasoning paths, more distributed evidence flow, and a more structured pattern of local connectivity, while failed predictions show shallower, fragmented, and overly concentrated evidence flow. Building on these findings, we develop a graph-based error detection framework that uses attribution-graph topology features. Furthermore, we show that attribution graphs enable targeted interventions. By reinforcing question-constrained evidence grounding, we reshape internal routing so that answer generation remains guided by the question, leading to more effective integration of retrieved information and fewer errors.

What does a book look like to a quantum computer? This paper takes eight classical works of the Renaissance and its late-antique inheritance -- from Augustine to Galileo -- and runs each through a neutral-atom quantum processor. The bridge is graphs: each textual unit becomes an atom, and graph edges are physical blockade constraints for engineered exact unit-disk designs, or a 2D approximation to the semantic graph for natural texts. Three contributions follow. First, we introduce rigidity rho, a metric for how unique a book's structural backbone is -- distinguishing Marguerite de Navarre's Heptameron (rigid, twelve-nouvelle hard core) from Boethius (fully fungible, every chapter substitutable). Second, we invert the pipeline: rather than extracting a graph from existing prose, we pick a target graph the hardware encodes natively, and write a book whose structure matches it. The twenty-nine texts written this way, collected under the name QOuLiPo, extend the OuLiPo tradition to graph-topological constraints and, together with the eight natural texts, form a benchmark distribution against which neutral-atom hardware can be tracked as it scales. Third, we run both natural and engineered texts on Pasqal's FRESNEL processor up to one hundred atoms; engineered texts reach high approximation ratios, the cleanest instances returning the exact backbone. A cloud-accessible quantum machine plus an agentic coding environment now lets a single investigator run this pipeline end-to-end. What is reported is an application layer, not a speedup -- humanistic instances ready to load onto neutral-atom processors as they scale, already complementing classical text analysis. The Digital Humanities community has a stake in building familiarity with this hardware now: the engineered-corpus design choices made today fix the benchmark distribution future hardware will be measured against.

Large language models (LLMs) often expose useful signals of self-monitoring: before solving a problem, they can estimate whether they are likely to succeed, and after solving it, they can judge whether their answer is likely to be correct. However, these signals are typically measured or elicited in isolation, rather than used to control inference. In this work, we ask whether LLMs possess latent metacognitive ability that can be turned into effective test-time control. Inspired by the Nelson--Narens theory from cognitive psychology, we propose a metacognitive harness that separates monitoring from reasoning. For each problem, the model first reports a pre-solve feeling-of-knowing (FOK) signal; after each solve attempt, it reports a post-solve judgment-of-learning (JOL) signal. Rather than treating these signals as passive confidence estimates, the harness turns them into an explicit control interface for reasoning: it decides when to trust the current solution, when to retry with compact metacognitive feedback, and when to pass multiple attempts to a final aggregator. Across text, code, and multimodal reasoning benchmarks, our harness substantially improves a fixed Claude Sonnet-4.6 base model without parameter updates or benchmark-specific fine-tuning. On the evaluated public benchmark snapshots, it raises pooled accuracy from 48.3 to 56.9 and exceeds the strongest listed leaderboard entries on the three primary evaluation settings: HLE-Verified, LiveCodeBench v6, and R-Bench-V. These results suggest that strong LLMs may already possess useful metacognitive ability, but require an explicit control harness to act on it during reasoning.

Long-horizon personalization requires dialogue assistants to retrieve user-specific facts from extended interaction histories. In practice, many relevant facts often have low semanticsimilarity to the query under dense retrieval. Standard Retrieval-Augmented Generation (RAG) and GraphRAG systems are still largely retrospective: they rely on embedding similarity to the query or on fixed graph traversals, so they often miss facts that matter for the user's needs but lie far from the query in embedding space. Inspired by prospection, the human ability to use imagined futures as cues for recall, we introduce Prospection-Guided Retrieval (PGR), which decouples retrieval from how memories are stored. Given a user query, PGR first expands the goal into a short Tree-of-Thought (ToT) or linear chain of plausible next steps, and uses these steps as retrieval probes rather than relying on the original query alone. The facts retrieved by these probes are then used to personalize the next round of prospection, enabling PGR to uncover additional memories that become relevant only after the simulation is grounded in the user's history. We also introduce MemoryQuest, a challenging multi-session benchmark in which each query is annotated with 3--5 dated reference facts subject to a low query-reference similarity constraint. Across 1,625 queries spanning 185 user profiles from 3 publicly available datasets, PGR-TOT substantially improves retrieval, including nearly 3x recall on MemoryQuest over the strongest baseline. In pairwise LLM-as-judge comparisons against baselines, PGR-generated responses are preferred on 89--98% of queries, with blinded human annotations on held-out subsets showing the same trend. Overall, the results demonstrate that explicit prospection yields large gains in long-horizon retrieval and response quality relative to similarity-only baselines.

In long conversations, an LLM can produce a next utterance that sounds plausible but rests on premises the conversation has already abandoned. Context-manipulation attacks against deployed agents now actively exploit this gap. We close it with a runtime verifier that maintains an explicit dependency graph: an LLM classifies each turn into one of 8 update operations drawn from four formalisms (dynamic epistemic logic, abductive reasoning, awareness logic, argumentation), and a symbolic engine records which claims depend on which evidence. Checking whether a continuation is supported reduces to a graph walk; retraction propagates through the same graph to flag exactly the conclusions that lose support, with linear per-turn cost and a formal conflict-free guarantee. On LongMemEval-KU oracle (n=78), the verifier reaches 89.7% accuracy vs. 88.5% for the LLM-only baseline (+1.3pp) and 87.2% for a transcript-RAG baseline matched on retrieval budget (+2.6pp); wins among disagreements are correct abstentions where the baseline confabulates. On LoCoMo's 60 official QA items the verifier is competitive with retrieval-augmented baselines. Beyond external benchmarks, we construct two multi-agent scenarios and a 50-item grounding test: on the 15-item stale-premise subset, the verifier reaches 100% accuracy vs. 93.3% (+6.7pp). These instantiate a soundness-faithfulness decomposition: the structural check is sound by construction, and per-deployment LLM extraction faithfulness is the empirical question we measure across four LLM families. The retraction check plateaus at microseconds while history-replay grows linearly with conversation length.

Channel state information (CSI) provides a widely available sensing modality for human and environment perception, but existing CSI sensing models usually rely on task-specific supervised training and require substantial labeled data for each task, device, user, or environment. This limits their scalability in practical deployments where unlabeled CSI is abundant but labeled data is costly to collect. In this paper, we present CSI-JEPA, a self-supervised predictive representation learning framework for label-efficient, multi-task Wi-Fi sensing. CSI-JEPA learns reusable temporal-spectral representations from unlabeled CSI samples by predicting latent features of masked channel regions from visible context. To better match the physical structure of CSI, CSI-JEPA tokenizes channel-response amplitude windows along the time and subcarrier dimensions. It then introduces a channel variation-aware masking strategy that samples predictive targets from regions with stronger local temporal and subcarrier-domain variations. After pretraining, the encoder is frozen and used as a backbone, with lightweight task-specific adapters added for downstream sensing tasks. We evaluate CSI-JEPA on seven real-world Wi-Fi sensing tasks spanning diverse objectives and deployment settings. The results show that CSI-JEPA improves downstream sensing performance over competitive baselines, achieving up to 10.64 percentage points mean accuracy gain over state-of-the-art supervised Transformer and matched-budget label savings of up to 98.0%.

Memory systems are critical for role-playing agents (RPAs) to maintain long-horizon consistency. However, existing RPA memory methods (e.g., profiling) mainly rely on recurrent summarization, whose compression inevitably discards important details. To address this issue, we propose a search-based memory framework called BOOKMARKS, which actively initializes, maintains, and updates task-relevant pieces of bookmarks for the current task (e.g., character acting). A bookmark is structured as the answer to a question at a specific point in the storyline. For each current task, BOOKMARKS selects reusable existing bookmarks or initializes new ones (at storyline beginning) with useful questions. These bookmarks are then synchronized to the current story point, with their answers updated accordingly, so they can be efficiently reused in future grounding rounds. Compared with recurrent summarization, BOOKMARKS offers (1) active grounding for capturing task-specific details and (2) passive updating to avoid unnecessary computation. In implementation, BOOKMARKS supports concept, behavior, and state searches, each powered by an efficient synchronization method. BOOKMARKS significantly outperforms RPA memory baselines on 85 characters from 16 artifacts, demonstrating the effectiveness of search-based memory for RPAs.

Learning of continuous exponential family distributions with unbounded support remains an important area of research for both theory and applications in high-dimensional statistics. In recent years, score matching has become a widely used method for learning exponential families with continuous variables due to its computational ease when compared against maximum likelihood estimation. However, theoretical understanding of the statistical properties of score matching is still lacking. In this work, we provide a non-asymptotic sample complexity analysis for learning the structure of exponential families of polynomials with score matching. The derived sample bounds show a polynomial dependence on the model dimension. These bounds are the first of its kind, as all prior work has shown only asymptotic bounds on the sample complexity.

Every AI benchmark operationalizes theoretical assumptions about the capability it claims to assess. When assumptions function as unexamined commitments, benchmarks stabilize the dominant paradigm by narrowing what counts as progress. Over time, narrow evaluation reorganizes capability concepts: architectures and definitions are selected for benchmark legibility until evaluation ceases to track an independent object and instead produces a version of the target defined by its own operational assumptions. The result is a trap: evaluation frameworks treat self-reinforcing assessments as valid, both creating and obscuring structural limits on what the current paradigm can accomplish. We introduce Epistematics, a methodology for deriving evaluation criteria directly from technical capability claims and auditing whether proposed benchmarks can discriminate the claimed capability from proxy behaviors. The contribution is meta-evaluative: an audit procedure, a failure mode taxonomy, and benchmark-design criteria for evaluating capability-evaluation coherence. We demonstrate the procedure through a worked audit of Dupoux et al. (2026), a proposal that revises the dominant paradigm's theoretical assumptions at the architectural level while reproducing them in its evaluation criteria, thereby entrenching the constraint it seeks to overcome in a form the evaluation cannot detect.

The primary way to establish and compare competencies in foundation and generative AI models has shifted from peer-reviewed literature to press releases and company blog posts, where model builders highlight results on selected benchmarks. These artifacts now largely define the state of the art for researchers and the public. Despite their prominence, which benchmarks model builders choose to highlight, and what they communicate through this selection, is underexamined. To investigate, we introduce and open-source Benchmarking-Cultures-25, a dataset of 231 benchmarks highlighted across 139 model releases in 2025 from 11 major AI builders, alongside an interactive tool to explore the data. Our analysis reveals a fragmented evaluation landscape with limited cross-model comparability: 63.2% of highlighted benchmarks are used by a single builder, and 38.5% appear in just one release. Few achieve widespread use (e.g., GPQA Diamond, LiveCodeBench, AIME 2025). Moreover, benchmarks are attributed different competencies by different builders, depending on their narrative. To disentangle these conflicting presentations, we develop a unified taxonomy mapping diverging terminology to a shared framework of measured signals based on what benchmark authors claim to measure. "General knowledge application" is the second most popular, yet vaguely defined, category. Qualitative analysis shows many such benchmarks deemphasize construct validity, instead framing results as indicators of progress toward AGI. Their authors claim to measure knowledge or reasoning broadly, yet mostly evaluate STEM subjects (especially math). We argue that highlighted benchmarks function less as standardized measurement tools and more as flexible narrative devices prioritizing market positioning over scientific evaluation. Data: https://hf.co/datasets/matybohacek/benchmarking-cultures-25; tool: https://bench-cultures.net.

Can a committee of weak reasoning-model calls reach the performance of much stronger models? We study verifier-backed committee search as inference-time boosting for reasoning language models. The mechanism is not simply that ``more agents help'': samples expose latent correct solutions, while critics and comparators must recover them without access to the hidden verifier. We formalize this view by separating proposal coverage, local identifiability, progress, and diversity. We prove that coverage can be amplified by repeated sampling, but cannot by itself create useful critics or comparators; reliable amplification requires an additional local soundness signal, such as execution, proof checking, type checking, tests, or constraint solving. We give rank-based bounds showing when local selection errors compose into reliable trajectories, and characterize the proposer-side ceiling: oracle best-of-\(k\) converges only to the mass of task slices on which the proposal system assigns nonzero useful probability. Empirically, on SWE-bench Verified, a single \texttt{GPT-5.4 nano} proposal solves \(67.0\%\) of tasks. Using the same nano model, our critic--comparator orchestration reaches \(76.4\%\) with \(k=8\) proposals, matching the standalone performance of \texttt{Gemini 3 Pro} and \texttt{Claude Opus 4.5} Thinking and approaching the \(79.0\%\) oracle best-of-\(8\) upper bound. Thus, many correct patches are already present in weak-model proposal pools; the main challenge is selecting them. The remaining failures are mostly proposal-coverage failures, indicating shared blind spots that stronger selection alone cannot close.

While generative models have shown promise in pediatric sleep analysis, the latent structure of their multimodal embeddings remains poorly understood. This work investigates session-wide diagnostic information contained in the sequences of 30-second pediatric PSG epochs embedded by a multimodal masked autoencoder. We test whether augmenting embeddings with PHATE-derived per-epoch coordinates and whole-night movement descriptors, persistent homology summaries of the embedding cloud, and EHR yields task-relevant signals. Simple linear and MLP models, chosen for interpretability rather than state-of-the-art performance, show that geometric, topological, and clinical features each provide complementary gains. For binary predictions, feature importance is task-dependent, and more expressive late-fusion models generally perform better, with AUPRC improving from 0.26 to 0.34 for desaturation, 0.31 to 0.48 for EEG arousal, 0.09 to 0.22 for hypopnea, and 0.05 to 0.14 for apnea. We also report Brier score and Expected Calibration Error, where the full fusion model yields the best calibration across all four binary tasks. Our study reveals that latent geometry/topology and EHR offer complementary, interpretable signals beyond embeddings, improving calibration and robustness under extreme imbalance.

Exploitation is not a binary event. It is a ladder of acquiring progressive capabilities, from executing a single buggy line of code to taking full control of the target. However, existing LLM security benchmarks treat a crash as exploitation success. That single binary outcome collapses the hard parts of exploitation: the transition from triggering a bug to constructing reusable primitives and control. We present ExploitBench, a capability-graded benchmark that decomposes exploitation into 16 measurable flags, from coverage and crash through sandbox primitives, arbitrary read/write, control-flow hijack, and arbitrary code execution. Each capability is verified by a deterministic oracle that uses a per-run randomized challenge-response for primitives, differential execution against ground-truth binaries to measure progress, and a signal-handler proof for code execution. We instantiate ExploitBench on 41 V8 bugs because V8 is both widely deployed and exploitation-hardened. We report three arms: <model,env> as the primary measurement of model-environment capability, <model,env, adaptive coaching> as a secondary arm that adds adaptive coaching to test whether targeted feedback shifts outcomes, and <model,env,harness> as an ablation that swaps in the model's native CLI to check whether vendor-side optimizations increase exploitation capabilities. Our results show a sharp capability split between publicly deployed frontier models and the private frontier. Across the 8 publicly deployed models tested, reaching the vulnerable code and triggering a crash is routine, but arbitrary code execution is not. The private model shows arbitrary code execution on approximately half. Overall, results suggest that exploit construction against hardened targets is an emerging frontier capability.

Safety evaluations for large language models (LLMs) increasingly target high-stakes National Security and Public Safety (NSPS) risks, yet multilingual safety is typically assessed through translation-only benchmarks that preserve the underlying scenario, and empirical evidence of how language and geopolitical context interact remains limited to a narrow set of language pairs. We introduce \emph{ROK-FORTRESS} https://huggingface.co/datasets/ScaleAI/ROK-FORTRESS_public, a bilingual, culturally adversarial NSPS benchmark that uses the English--Korean language pair and U.S.--ROK geopolitical axis as a case study, separating the effects of language and geopolitical grounding via a \emph{transcreation matrix}: adversarial intents are evaluated under controlled combinations of (i) English versus Korean language and (ii) U.S.\ versus Korean entities, institutions, and operational details. Each adversarial prompt is paired with a dual-use benign counterpart to quantify over-refusal. Model responses are then scored using calibrated LLM-as-a-judge panels, applying our expert-crafted, prompt-specific binary rubrics. Across a dual-track set of frontier and Korean-optimized models, we find a consistent suppression effect in Korean variants and substantial model-to-model variation in how geopolitical grounding interacts with language. In many models, Korean grounding mitigates the Korean language-driven suppression -- with no model showing significant amplification in the other direction -- indicating that, at least in the English--Korean case, safety behavior is shaped by language-as-risk signals and context interactions that translation-only evaluations miss. The transcreation matrix methodology is designed to generalize to other language--culture pairs.

Objective: The primary goal of this study was to systematically examine the impact of commonly used imbalance handling methods (IHMs) on predictive performance in biomedical binary classification, considering the interplay between model complexity and diverse data modalities. Material and Methods: We evaluated five representative IHMs: random undersampling (RUS), random oversampling (ROS), SMOTE, re-weighting (RW), and direct F1-score optimization (DMO), against a raw training (RAW) baseline. The evaluation encompassed three public biomedical datasets: MIMIC-III (tabular), ADE-Corpus-V2 (text), and MURA (image), spanning three common biomedical data modalities. To assess varying model complexity, we employed a range of architectures, from classical logistic regression and random forest to deep neural networks, including multilayer perceptron (MLP), BiLSTM, BERT, DenseNet, and DINOv2. Results: For simpler models such as logistic regression on tabular data, IHMs yielded no significant advantage over the RAW baseline, aligning with prior findings. However, clear benefits were observed for more complex models and unstructured data: (a) ROS and RW consistently enhanced the performance of powerful models; (b) direct F1-score optimization demonstrated utility primarily for unstructured text and image data; and (c) RUS and SMOTE consistently degraded performance and are therefore not recommended. Conclusion: The effectiveness of IHMs depends on both model complexity and data modality. Performance gains are most pronounced when leveraging appropriate IHMs, such as ROS, RW, and DMO, on high-complexity models.

bde is a user-friendly Python package for Bayesian Deep Ensembles with a particular focus on tabular data. Built on an efficient JAX implementation of the sampling-based inference method Microcanonical Langevin Ensembles (MILE), it provides scikit-learn compatible estimators for fast training, efficient Markov Chain Monte Carlo sampling, and uncertainty quantification in both regression and classification tasks.

Integration against a probability distribution given its unnormalized density is a central task in Bayesian inference and other fields. We introduce new methods for approximating such expectations with a small set of weighted samples -- i.e., a quadrature rule -- constructed via an interacting particle system that minimizes maximum mean discrepancy (MMD) to the target distribution. These methods extend the classical mean shift algorithm, as well as recent algorithms for optimal quantization of empirical distributions, to the case of continuous distributions. Crucially, our approach creates dynamics for MMD minimization that are invariant to the unknown normalizing constant; they also admit both gradient-free and gradient-informed implementations. The resulting mean shift interacting particle systems converge quickly, capture anisotropy and multi-modality, avoid mode collapse, and scale to high dimensions. We demonstrate their performance on a wide range of benchmark sampling problems, including multi-modal mixtures, Bayesian hierarchical models, PDE-constrained inverse problems, and beyond.

We study learning when the learned object is executable solver code rather than a predictor. In this setting, correctness is not enough: two solvers may both return valid solutions on the deployment distribution while differing substantially in runtime. Given samples from an unknown task distribution, the learner returns code evaluated on fresh instances by both solution quality and execution time. Our central abstraction is a \emph{solver hint}: reusable structure inferred from samples and compiled into specialized solver code. We prove that the empirically fastest sample-consistent solver from a fixed library generalizes in both correctness and runtime, and that statistically identifiable hints can be recovered and compiled from polynomially many samples. Empirically, we instantiate the framework with LLM code agents on \(21\) structured combinatorial-optimization target distributions across seven problem classes. The synthesized solvers reach mean normalized quality \(0.971\), improve by \(+0.224\) over the average heuristic pool and by \(+0.098\) over the highest-quality heuristic, and are \(336.9\times\), \(342.8\times\), and \(16.1\times\) faster than the quality-best heuristic, Gurobi, and the selected time-limited exact backend, respectively. On released PACE 2025 Dominating Set private instances, the synthesized solver is valid on all \(100\) graphs and runs about two orders of magnitude faster than top competition solvers, with a moderate quality gap. Inspection shows that many gains come from changing the computational scale: replacing ambient exponential search or general-purpose optimization with compiled distribution-specific computation.

Interactive agent benchmarks face a tension between scalable construction and realistic workflow evaluation. Hand-authored tasks are expensive to extend and revise, while static prompt evaluation misses failures that only appear when agents operate over persistent state. Existing interactive benchmarks have advanced agent evaluation significantly, but most initialize tasks from clean state and do not systematically test how agents handle pre-existing partial, stale, or conflicting artifacts. We present \textbf{ClawForge}, a generator-backed benchmark framework for executable command-line workflows under state conflict. The framework compiles scenario templates, grounded slots, initialized state, reference trajectories, and validators into reproducible task specifications, and evaluates agents step by step over persistent workflow surfaces using normalized end state and observable side effects rather than exact trajectory matching. We instantiate this framework as the ClawForge-Bench (17 scenarios, 6 ability categories). Results across seven frontier models show that the best model reaches only 45.3% strict accuracy, wrong-state replacement remains below 17\% for all models, and the widest model separation (17% to 90%) is driven by whether agents inspect existing state before acting. Partial-credit and step-efficiency analyses further reveal that many failures are near-miss closures rather than early breakdowns, and that models exhibit qualitatively different failure styles under state conflict.

Fast Healthcare Interoperability Resources (FHIR) is the dominant standard for interoperable exchange of healthcare data. In FHIR, electronic health records form a directed graph of resources. Answering clinically meaningful questions over FHIR requires agents to perform multi-step reasoning, filtering, and aggregation across multiple resource types. Prior work shows that even tool-augmented LLM agents (retrieval, code execution, multi-turn planning) often select the wrong resources or violate traversal constraints. We study this problem in the context of FHIR-AgentBench, a benchmark for realistic question answering over real-world hospital data, and frame reasoning on FHIR as a sequential decision-making problem over a queryable structured graph. We implement a multi-turn CodeAct agent and post-train it with reinforcement learning using a custom harness and tools. A LLM Judge provides execution-grounded rewards. Compared to prompt-based, closed-model baselines, RL post-training improves performance while enforcing data-integrity constraints. Empirically, our approach improves answer correctness from 50% (o4-mini) to 77% on FHIR-AgentBench using a smaller and cheaper Qwen3-8B model. We present an end-to-end post-training pipeline (environment building, harness construction, model training and custom evaluation) that reliably improves multi-turn reasoning over structured clinical graphs.

How do artificial neural networks bind concepts to form complex semantic structures? Here, we propose a simple neural code, whereby the existence and the type of relations between entities are represented by the distance and the direction between their embeddings, respectively. We test this hypothesis in a variety of Large Language Models (LLMs), each input with natural-language descriptions of minimalist tasks from five different domains: arithmetic, visual scenes, family trees, metro maps and social interactions. Results show that the true semantic structures can be linearly recovered with a Polar Probe targeting a subspace of LLMs' layer activations. Second, this code emerges mostly in middle layers and improves with LLM performance. Third, these Polar Probes successfully generalize to new entities and relation types, but degrades with the size of the semantic structure. Finally, the quality of the polar representation correlates with the LLM's ability to answer questions about the semantic structure. Together, these findings suggest that LLMs learn to build complex semantic structures by binding representations with a simple geometrical principle.

Geospatial foundation models compress multispectral observations into dense embeddings increasingly used in natural-language environmental reasoning systems. A single planetary-scale model, e.g. Google AlphaEarth, handles broad characterization well but may compromise on specialized hydrologic signals. Such generalist models are also often inaccessible, expensive, and require large-scale compute. We propose Mini-JEPAs: a fleet of small sensor-specialized Joint Embedding Predictive Architecture (JEPA) foundation models consulted by a routing agent for specialized questions. We pretrained five 22M-parameter Mini-JEPAs sharing an identical Vision Transformer backbone, JEPA recipe, and 64-d output space, using Sentinel-2 optical, Sentinel-1 SAR, MODIS thermal, multi-temporal Sentinel-2 phenology, and a topography-soil stack. Each Mini-JEPA reconstructs the variable matched to its sensor, with cross-validated $R^2$ reaching 0.97 for elevation, 0.97 for temperature, and 0.81 for precipitation. The five manifolds differ in geometric structure, with global participation ratios from 8.9 to 20.2 and local intrinsic dimensionalities from 2.3 to 9.0. Joint topography-soil and phenology models add predictive value beyond AlphaEarth alone for soil moisture, aridity, and precipitation ($ΔR^2$ up to 0.031). A router LLM reads per-modality references and selects appropriate sensors with a perfect hit rate over a curated question set. In paired LLM-as-Judge evaluation, dual retrieval over AlphaEarth and the routed fleet outperforms AlphaEarth alone on physics-matched questions (Cohen's $d = 1.10$, $p = 0.031$). Locally-trained Mini-JEPAs can be operationalized for hydrologic intelligence with modest compute.

An AI system for professional floor plan design must precisely control room dimensions and areas while respecting the desired connectivity between rooms and maintaining functional and aesthetic quality. Existing generative approaches focus primarily on respecting the requested connectivity between rooms, but do not support generating floor plans that respect numerical constraints. We introduce a text-based floor plan generation approach that fine-tunes a large language model (LLM) on real plans and then applies reinforcement learning with verifiable rewards (RLVR) to improve adherence to topological and numerical constraints while discouraging invalid or overlapping outputs. Furthermore, we design a set of constraint adherence metrics to systematically measure how generated floor plans align with user-defined constraints. Our model generates floor plans that satisfy user-defined connectivity and numerical constraints and outperforms existing methods on Realism, Compatibility, and Diversity metrics. Across all tasks, our approach achieves at least a 94% relative reduction in Compatibility compared with existing methods. Our results demonstrate that LLMs can effectively handle constraints in this setting, suggesting broader applications for text-based generative modeling.

Biomedical retrieval-augmented large language models (LLMs) often face evidence that is incomplete, misleading, or internally contradictory, yet evaluation usually emphasizes answer accuracy under helpful context rather than reliability under conflict. Using HealthContradict, we evaluate six open-weight LLMs under five controlled evidence conditions: no retrieved context, correct-only context, incorrect-only context, and two mixed conditions containing both correct and contradictory documents in opposite orders. In this conflicting-evidence order contrast, where the same two documents are both present and only their order is reversed, accuracy drops for every model and 11.4%--25.2% of predictions flip. To support abstention in these difficult cases, we also evaluate a conflict-aware abstention score that combines model confidence with a detector of evidence conflict. In the two hardest conditions, this score improves selective accuracy over confidence-only, with mean gains of 7.2--33.4 points in incorrect-only (`IC') and 3.6--14.4 points in incorrect-first conflicting (`ICC') conditions across 75%, 50%, and 25% coverage. These results show that conflicting biomedical evidence is both an uncertainty and robustness problem and motivate evaluation and abstention methods that explicitly account for evidence disagreement.

While interpretable prototype networks offer compelling case-based reasoning for clinical diagnostics, their raw continuous outputs lack the semantic structure required for medical documentation. Bridging this gap via standard Retrieval-Augmented Generation (RAG) routinely triggers ``retrieval sycophancy,'' where Large Language Models (LLMs) hallucinate post-hoc rationalizations to align with visual predictions. We introduce ProtoMedAgent, a framework that formalizes multimodal clinical reporting as an iterative, zero-gradient test-time optimization problem over a strict neuro-symbolic bottleneck. Operating on a frozen prototype backbone, we distill latent visual and tabular features into a discrete semantic memory. Online generation is strictly constrained by exact set-theoretic differentials and a reflective Scribe-Critic loop, mathematically precluding unsupported narrative claims. To safely bound data disclosure, we introduce a semantic privacy gate governed by $k$-anonymity and $\ell$-diversity. Evaluated on a 4,160-patient clinical cohort, ProtoMedAgent achieves 91.2\% Comparison Set Faithfulness where it fundamentally outperforms standard RAG (46.2\%). ProtoMedAgent additionally leverages a binding $\ell$-diversity phase transition to systematically reduce artifact-level membership inference risks by an absolute 9.8\%.

Hospital pharmacists make high-stakes decisions to mitigate drug shortages under uncertainty, time pressure, and patient risk. Interviews revealed that pharmacists focus attention on a small subset of drugs, limiting cognitive effort to the most urgent cases. Motivated by these findings, we formalize a bounded-rational, attention-guided decision framework that dynamically decomposes drugs into a subset for high-cost reasoning and a complementary subset for low-cost monitoring. We develop two agents: an Expert Agent that applies attention weights derived from pharmacist interviews, and a Learner Agent that adapts attention allocation over time through experience. Across simulated scenarios spanning short to long horizons, we show that attention-guided planning supports stable decision-making without complete state reasoning. These results suggest that a primary decision is not what action to take, but where to allocate cognitive effort, and that attention-guided, satisficing strategies can reduce problem complexity while maintaining stable performance.

Diabetic Retinopathy (DR) is one of the leading causes of preventable blindness, yet rural regions often lack the specialists and infrastructure needed for early detection. Although cloud-based deep learning systems offer high accuracy, they face significant challenges in these settings due to high latency, limited bandwidth, and high data transmission costs. To address these challenges, we propose a two-tier edge-cloud cascade on the public APTOS 2019 Blindness Detection dataset. Tier 1 runs a lightweight MobileNetV3-small model on a local clinic device to perform a binary triage between Referable DR (Classes 2-4) and Non-referable DR (Classes 0-1). Tier 2 runs a RETFoundDINOv2 model in the cloud for ordinal severity grading, but only on the subset of images flagged as referable by Tier 1. On a stratified APTOS test split of 733 images, Tier 1 reaches 98.99% sensitivity and 84.37% specificity at a validation-tuned high-sensitivity threshold. The default cascade forwards 49.52% of test images to Tier 2, reducing cloud calls by 50.48% relative to using a cloud-based model for all images. In the deployed 4-class output space (Class 0-1 / Class 2 / Class 3 / Class 4), the cascade obtains 80.49% accuracy and 0.8167 quadratic weighted kappa; the cloud-only baseline obtains 80.76% accuracy and 0.8184 quadratic weighted kappa. On APTOS, the cascade cuts cloud use by about half with a modest drop in grading performance. Index Terms: Diabetic Retinopathy, Edge-Cloud Cascade, MobileNetV3-small, RETFound-DINOv2, Retinal Screening, tele-ophthalmology

Autonomous language-model agents increasingly combine planning, tool use, document processing, browsing, code execution, and verification loops. These capabilities make agent systems more useful, but they also introduce operational failure modes that are not visible from final accuracy alone. This report presents ChromaFlow, a tool-augmented autonomous reasoning framework built around planner-directed execution, specialized tool use, and telemetry-driven evaluation. We analyze ChromaFlow on GAIA 2023 Level-1 validation tasks under clean evaluation constraints. A frozen full Level-1 baseline achieved 29/53 correct answers, or 54.72%. A later recovery configuration with expanded orchestration achieved 27/53 correct answers, or 50.94%, while increasing tracebacks, timeout events, tool-failure mentions, token-line calls, and campaign-log cost estimates. Two randomized 20-task smoke evaluations produced 12/20 and 11/20 correct answers, showing that small diagnostic gains can be unstable across samples. The central result is therefore a negative ablation: more aggressive orchestration did not improve full-set performance and increased operational noise. The report argues that bounded planner escalation, deterministic extraction, evidence reconciliation, and explicit run gates should be treated as first-order requirements for reliable autonomous agent evaluation.

Chain-of-thought (CoT) reasoning with self-consistency improves performance by aggregating multiple sampled reasoning paths. In this setting, correctness is no longer tied to a single reasoning trace but to the aggregation rule over a pool of candidate paths, making aggregation uncertainty the central challenge. This issue is critical where confidently incorrect answers are far more costly than abstentions. We introduce a conformal procedure for CoT reasoning that directly addresses aggregation uncertainty. Our approach replaces majority voting with weighted score aggregation over reasoning paths and calibrates an abstention rule using conformal risk control. This approach leads to finite-sample guarantees on the confident-error rate--the probability that the system answers and is wrong. We further identify score separability as the key condition under which abstention provably improves selective accuracy, and derive closed-form expressions that predict accuracy gains from calibration data alone. The method is fully inference-time, and requires no retraining. Across four benchmarks, four open-source models, and three score classes, realized confident-error rates are consistent with the prescribed targets up to calibration-split and test-set variability. Our method achieves $90.1\%$ selective accuracy on GSM8K by abstaining on less than $5\%$ of problems, compared with $82\%$ accuracy under majority-voting baseline.

We put forth a critical theoretical framework for analyzing generative models both descriptively and normatively. Our thesis is that generative models automate the production not only of intellectual labor or intelligence, but of a broader set of human social capacities we name "social doing." We do this by historicizing the commodification of sociality in the digital economy, leading to the availability of social data as the precondition for generative models. We elaborate our definition of "social doing" by drawing a distinction between "use" and "exchange" sociality and further differentiate between the ways that generative models either substitute for or mediate existing social relations and processes. We then turn to existing empirical research on how people use generative model-based products and the effects that their use has upon them. In this, we introduce the concept of Synthetic Sociality, a social reality in part fabricated by Silicon Valley's privately owned and undemocratically governed generative models. Lastly, we offer a normative analysis based on our findings and framework, and discuss future design opportunities.

In recent years, a variety of powerful LLM-based agentic systems have been applied to automate complex tasks through task orchestration. However, existing orchestration methods still face key challenges, including strategy collapse under reward maximization, high gradient variance with opaque credit assignment, and unguided skill evolution whose decisions are typically made by directly prompting an LLM to judge rather than derived from principled training signals. To address these challenges, we propose SkillFlow, a flow-based framework that takes a trainable Supervisor as the agent and a structured environment with dynamic skill library and frozen executor, automating task orchestration through multi-turn interaction. SkillFlow employs Tempered Trajectory Balance (TTB), a regression-based flow-matching loss that samples trajectories proportional to reward, preserving diverse orchestration strategies rather than collapsing to a single mode. The same flow objective yields a jointly learned backward policy that provides transparent per-step credit assignment at zero additional inference cost. Building on these flow diagnostics, a recursive skill evolution mechanism determines when to evolve, what skills to create or prune, and where decision gaps lie -- closing the loop from training signal to autonomous capability growth. Experimental results on 14 datasets show that SkillFlow significantly outperforms baselines across question answering, mathematical reasoning, code generation, and real-world interactive decision making tasks. Our code is available at https://anonymous.4open.science/r/SkillFlow-E850.

Large Language Models (LLMs), when trained on web-scale corpora, inherently absorb toxic patterns from their training data. This leads to ``toxic degeneration'' where even innocuous prompts can trigger harmful outputs. This phenomenon poses significant risks for real-world deployments. Thus, necessitating effective mitigation strategies that should maintain model utility while ensuring safety. In this comprehensive replication study, we evaluate the efficacy of \textbf{DExperts} (Decoding-time Experts), which is an inference-time mitigation technique that steers generation without requiring model retraining. We structured our research into three systematic phases: (1) establishing baseline toxicity measurements using \textbf{RealToxicityPrompts} on standard GPT-2 models; then (2) implementing and evaluating DExperts to mitigate explicit toxicity; and finally (3) stress-testing the method against implicit hate speech using the adversarial \textbf{ToxiGen} dataset. Our empirical results confirm that while DExperts achieves near-perfect safety rates (100\%) on explicit toxicity benchmarks, it exhibits brittleness against adversarial, implicit hate speech, with safety rates dropping to 98.5\%. Furthermore, we quantify a critical trade-off. The method introduces a $\sim$10x latency penalty (from 0.2s to 2.0s per generation), posing challenges for real-time deployment scenarios. This study contributes to the growing body of work on AI safety by highlighting the robustness gap between explicit and implicit toxicity mitigation. We emphasize the need for more sophisticated approaches that generalize across diverse hate speech patterns without prohibitive computational costs.

Code agents must both reason over long-horizon repository state and obey strict tool-use protocols. In paired Instruct/Thinking checkpoints, these capabilities are complementary but misaligned. The Instruct model is concise and tool-disciplined, whereas the Thinking model offers stronger planning and recovery behavior but often over-deliberates and degrades agent performance. We present CRANE (Constrained Reasoning Injection for Code Agents via Nullspace Editing), a training-free parameter-editing method that treats the Thinking-Instruct delta as a directional pool of candidate reasoning edits for the Instruct backbone. CRANE combines magnitude thresholding to denoise the delta, a Conservative Taylor Gate to retain edits that are jointly beneficial for reasoning transfer and tool-use preservation, and Graduated Sigmoidal Projection to suppress format-critical update directions. By merging paired Instruct and Thinking checkpoints, CRANE delivers strong gains over either individual model while preserving Instruct-level efficiency: on Roo-Eval it achieves pass1 of 66.2% (+19.5%) for Qwen3-30B-A3B and 81.5% (+8.7%) for Qwen3-Next-80B-A3B; on SWE-bench-Verified it resolves up to 14 additional instances at both scales (122/500 and 180/500); and on Terminal-Bench v2 it improves pass1/pass5 by up to 2.3%/7.8%, reaching 7.6%/17.9% and 14.8%/30.3%, respectively, consistently outperforming alternative merging strategies across all three benchmarks.

Large language models (LLMs) have revolutionized natural language processing. Understanding their internal mechanisms is crucial for developing more interpretable and optimized architectures. Mechanistic interpretability has led to the development of various methods for assessing layer relevance, with cosine similarity being a widely used tool in the field. On this work, we demonstrate that cosine similarity is a poor proxy for the actual performance degradation caused by layer removal. Our theoretical analysis shows that a layer can exhibit an arbitrarily low cosine similarity score while still being crucial to the model's performance. On the other hand, empirical evidence from a range of LLMs confirms that the correlation between cosine similarity and actual performance degradation is often weak or moderate, leading to misleading interpretations of a transformer's internal mechanisms. We propose a more robust metric for assessing layer relevance: the actual drop in model accuracy resulting from the removal of a layer. Even though it is a computationally costly metric, this approach offers a more accurate picture of layer importance, allowing for more informed pruning strategies and lightweight models. Our findings have significant implications for the development of interpretable LLMs and highlight the need to move beyond cosine similarity in assessing layer relevance.

Fairness in toxicity classification involves three integrated axes: ranking, calibration, and abstention. Training-time interventions and post-hoc safety mechanisms cannot be evaluated independently because the former determines the efficacy of the latter. We compare Empirical Risk Minimization (ERM), instance-level reweighting, and Group DRO across these axes, combined with temperature scaling, confidence-based abstention, and per-identity threshold optimization. Evaluation uses subgroup AUC, BPSN/BNSP AUC, error gaps, and per-subgroup Expected Calibration Error (ECE) with bootstrap CIs ($n = 1000$). We report four findings. (1) Calibration disparity is a hidden fairness violation. ERM has near-perfect aggregate calibration ($0.013$) but is significantly miscalibrated across all identity subgroups ($+0.029$ to $+0.134$). (2) Training interventions reshape rather than eliminate disparity. Reweighted ERM improves ranking (BPSN AUC $+0.06$ to $+0.12$) but worsens the calibration-fairness gap by up to $+0.232$. Group DRO eliminates calibration disparity but only by becoming uniformly miscalibrated globally (ECE $0.118$). (3) Post-hoc methods inherit training failure modes. Temperature scaling fails because miscalibration is non-uniform. Confidence-based abstention works under ERM but breaks under DRO, where the risk-coverage curve rises with deferral. (4) Abstention itself is unfair. Confidence-based deferral helps background content far more than identity-mentioning content. We argue that SRAI fairness requires a multi-axis framework: methods that differ only in aggregate ranking can differ sharply in failure modes that determine real-world harm.

Deep neural networks have achieved strong performance in genomic sequence classification; however, relating their predictions to biologically meaningful sequence patterns remains challenging. In this work, we present AttnGen, an attention-guided training framework that embeds interpretability directly into the optimization process. AttnGen computes nucleotide-level importance scores using an attention mechanism and progressively suppresses low-contribution positions during training. This encourages the model to focus its predictions on a compact set of informative regions while reducing reliance on noisy sequence elements. We evaluate AttnGen on the standardized demo_human_or_worm benchmark, a binary classification task over 200-nucleotide sequences. With moderate masking, AttnGen achieves a validation accuracy of 96.73%, outperforming a conventional CNN baseline with 95.83% accuracy, while also exhibiting faster convergence and improved training stability. To assess whether the learned importance scores reflect functionally relevant signal, we conduct perturbation-based analysis by removing high-saliency nucleotides. This causes accuracy to drop from 96.9% to near chance level on a 3,000-sequence evaluation set, indicating that the model relies on a relatively small subset of informative positions. Our analysis shows that masking 10--20% of positions provides the most favorable trade-off between predictive performance and interpretability. These results suggest that attention-guided masking not only improves classification performance but also reshapes how models distribute importance across sequence positions. Although this study focuses on short genomic sequences, the proposed approach may extend to more complex interpretable sequence modeling settings.

Distilling reasoning traces from strong large language models into smaller ones is a promising route to improve intelligence in resource-constrained settings. Existing approaches face a fundamental trade-off: offline distillation from teacher-generated traces provides high-quality, sample-efficient supervision but suffers from distributional drift: during training, the student model conditions on teacher-generated prefixes, whereas during inference the student autoregresses on self-generated prefixes, leading to compounding errors over long reasoning trajectories. Meanwhile, on-policy or self-distillation methods better match the student's inference-time distribution, but require costly online sampling and often produce low-quality traces in early training. We propose a principled offline reasoning distillation framework that preserves the efficiency and supervision quality of offline teacher-generated data while correcting teacher-student distribution drift. It adaptively emphasizes teacher supervision that is better aligned with the student's on-policy distribution. Evaluations on mathematical reasoning benchmarks of GSM8K, MATH, MATH500, and harder held-out competition-style tasks, including AMC, AIME, and OlympiadBench, show that our method improves reasoning accuracy over prior offline distillation algorithms and yields more stable reasoning traces while preserving instruction-following capabilities. Our work shows that lightweight, distribution-correction-aware training can substantially strengthen offline reasoning distillation without online rollouts.

Irregularly sampled multivariate event streams remain a stubbornly difficult modality for generative modeling: tokenization-based approaches break down when inter-event intervals vary by orders of magnitude, and neural temporal point processes are bottlenecked by window-level numerical quadrature. We (i) propose SurF, a generative model that uses the Time Rescaling Theorem (TRT) as a learnable bijection between event sequences and i.i.d.\ unit-rate exponential noise, enabling a single model to be trained across heterogeneous event-stream datasets; (ii) three efficient parameterizations of the cumulative intensity that scale to long sequences; and (iii) a Transformer-based encoder for multi-dataset pretraining. On six real-world benchmarks, SurF achieves the best reported time RMSE on Earthquake, Retweet, and Taobao, and is within trial-level noise of the strongest specialist on the remaining three. Under a strict leave-one-out protocol, the held-out checkpoint beats every classical and neural-autoregressive baseline on 5/6 datasets and beats every baseline on Amazon and Earthquake, an initial step toward foundation models over asynchronous event streams.

We introduce CurveBench, a benchmark for hierarchical topological reasoning from visual input. CurveBench consists of \textbf{756 images} of pairwise non-intersecting Jordan curves across easy, polygonal, topographic-inspired, maze-like, and dense counting configurations. Each image is annotated with a rooted tree encoding the containment relations between planar regions. We formulate the task as structured prediction: given an image, a model must recover the full rooted containment tree induced by the curves. Despite the visual simplicity of the task, the strongest evaluated model, Gemini 3.1 Pro, achieves only \textbf{71.1\%} tree-generation accuracy on CurveBench-Easy and \textbf{19.1\%} on CurveBench-Hard. We further demonstrate benchmark utility through RLVR-style fine-tuning of open-weight vision-language models. Our trained Qwen3-VL-8B model improves over \texttt{Qwen-3-VL-8B-Thinking} from \textbf{2.8\%} to \textbf{33.3\%} tree-generation accuracy on CurveBench-Easy, exceeding GPT-5.4 and Claude Opus 4.5 under our evaluation protocol. The remaining gap, especially on CurveBench-Hard, shows that exact topology-aware visual reasoning remains far from solved.

Financial distress prediction remains a significant challenge in enterprise risk analysis due to the highly imbalanced nature of real-world financial datasets, where bankrupt or distressed firms typically constitute only a small minority of observations. This paper presents a comparative evaluation of classical statistical methods, ensemble learning approaches, and exploratory neural models for minority-class financial distress prediction under class imbalance constraints. The study incorporates structured preprocessing, imbalance mitigation using the Synthetic Minority Oversampling Technique (SMOTE), comparative evaluation across ensemble learning architectures including XGBoost, CatBoost, LightGBM, Random Forest, and explainability analysis using SHAP-based feature attribution methods. Experimental evaluation demonstrates that gradient-boosting approaches achieved improved minority-class sensitivity relative to baseline statistical classifiers under severe imbalance conditions. The workflow additionally emphasises reproducibility, interpretability, auditability, and governance-oriented machine learning evaluation within enterprise financial risk environments. The work is positioned as an applied engineering evaluation intended to support reproducible and interpretable machine learning workflows for financial distress prediction under severe class imbalance constraints.

Early-stage Parkinson's disease (EarlyPD) detection from speech is clinically meaningful yet underexplored, and published results are hard to compare because studies differ in datasets, languages, tasks, evaluation protocols, and EarlyPD definitions. To address this issue, we propose the first benchmark for speech-based EarlyPD detection, with a speaker-independent split designed for fair and replicable cross-method evaluation on researcher-accessible datasets. The benchmark covers three common speech tasks and evaluates methods under different training-resource settings. We also present multi-dimensional evaluation breakdowns by dataset, aggregation level, gender, and disease stage to support fine-grained comparisons and clinical adoption. Our results provide a replicable reference and actionable insights, encouraging the adoption of this publicly available benchmark to advance robust and clinically meaningful EarlyPD detection from speech.

Continual test-time adaptation (CTTA) updates a pretrained model online on an unlabeled, non-stationary stream while anchoring it to a frozen source checkpoint. This anchor is useful only when the source remains reliable. On CCC-Hard, however, a ResNet-50 source falls to approximately $1.3\%$ top-$1$ accuracy, while existing source-anchored CTTA methods continue applying the same anchor strength. We call this failure mode blind anchoring and propose RMemSafe, a reliability-gated extension of ROID that uses the frozen source's normalized predictive entropy to attenuate all explicit source-coupled uses in the objective. When the source posterior approaches uniformity, the gate closes: the source anchor and agreement filter vanish, and the objective reduces to a source-agnostic fallback comprising ROID's base losses plus marginal calibration. Combined with ASR, RMemSafe achieves the lowest error on $8$ of $9$ matched-split continual-corruption cells and is the best reset-based method on all $9$, improving ROID+ASR by $1.05$~pp on ResNet-50 and $0.48$~pp on ViT-B/16. A controlled source-degradation sweep shows a $1.13{\times}$ shallower harm slope than ROID+ASR, consistent with the graceful-decay prediction. The entropy gate detects high-entropy source collapse, not confidently wrong low-entropy sources; this scope is explicitly evaluated and discussed.

While synthetic data generation with large language models (LLMs) is widely used in post-training pipelines, existing approaches typically generate full outputs before applying quality filters, leading to substantial token waste on samples that are ultimately discarded. To address this, we propose Multi-Stage In-Flight Rejection (MSIFR), a lightweight, training-free framework that detects and terminates low-quality generation trajectories at intermediate checkpoints before they reach full completion. MSIFR decomposes the generation process into sequential stages and applies fast rule-based validators to identify arithmetic inconsistencies, hallucination patterns, and formatting violations, enabling early rejection of faulty samples. We formalize in-flight rejection as a sequential decision process and show that any non-trivial discard policy reduces expected token consumption, with stage-wise savings increasing when rejection occurs earlier in the generation pipeline. We further demonstrate that conditional utility estimates form a martingale, ensuring that early, in-flight rejection does not bias the expected utility of retained samples. Across five instruction-tuned models and seven reasoning benchmarks, MSIFR reduces token consumption by 11%-77% as a standalone method, and up to 78.2% when combined with early-exit methods, while preserving or improving evaluation accuracy. These results confirm that MSIFR provides a practical mechanism for improving the efficiency of LLM-based synthetic data generation without additional training or architectural changes.

Current autoformalization benchmarks are largely focused on olympiad or undergraduate mathematics, while graduate and research-level mathematics remains underexplored. In this paper, we introduce MathAtlas, the first large-scale autoformalization benchmark of in the wild graduate-level mathematics, containing ~52k theorems, definitions, exercises, examples, and proofs extracted from 103 graduate mathematics textbooks. MathAtlas is enriched with a mathematical dependency graph containing ~178k relations, and is the first autoformalization benchmark to include such relations, facilitating evaluation and development of dependency-aware autoformalization systems. Our extensive experiments show that MathAtlas is high quality but extremely challenging: strong baselines achieve at most 9.8% correctness on theorem statements and 16.7% on definitions. Furthermore, we find performance of state-of-the-art models degrades substantially with dependency depth: on MA-Hard, a subset of 700 entities with the deepest dependency trees, the best model achieves only 2.6% correctness for autoformalization on this challenging dataset. We release MathAtlas to the community as a benchmark set for large-scale autoformalization of graduate-level mathematics in the wild.

Most existing dialogue systems are user-driven, primarily designed to fulfill user requests. However, in many critical real-world scenarios, a conversational agent must proactively extract information to achieve its own objectives rather than merely respond. To address this gap, we introduce \emph{Inquisitive Conversational Agents (ICAs)} and develop an ICA specifically tailored to U.S. Supreme Court oral arguments. We propose a Dual Hierarchical Reinforcement Learning framework featuring two cooperating RL agents, each with its own policy, to coordinate strategic dialogue management and fine-grained utterance generation. By learning when and how to ask probing questions, the agent emulates judicial questioning patterns and systematically uncovers crucial information to fulfill its legal objectives. Evaluations on a U.S. Supreme Court dataset show that our method outperforms various baselines across multiple metrics. It represents an important first step toward broader high-stakes, domain-specific applications.

Parameter-Efficient Fine-Tuning (PEFT) is widely used for adapting Large Language Models (LLMs) for various tasks. Recently, there has been an increasing demand for fine-tuning a single LLM for multiple tasks because it requires overall less data for fine-tuning thanks to the common features shared among tasks. More importantly, LLMs are resource demanding and deploying a single model for multiple tasks facilitates resource consolidation and consumes significantly less resources compared to deploying individual large model for each task. Existing PEFT methods like LoRA and Prefix Tuning are designed to adapt LLMs to a specific task. LoRA and its variation focus on aligning the model itself for tasks, overlooking the importance of prompt tuning in multi-task learning while Prefix Tuning only adopts a simple architecture to optimize prompts, which limits the adaption capabilities for multi-task. To enable efficient fine-tuning for multi-task learning, it is important to co-optimize prompt optimization and model adaptation. In this work, we propose a Parameter-Efficient Multi-task Learning (\PM), which employs a neural architecture engineering method for optimizing the continuous prompts while also performing low-rank adaption for model weights. We prototype PEML by creating an automated framework for optimizing the continuous prompts and adapting model weights. We evaluate PEML against state-of-the-arts multi-task learning methods MTL-LoRA, MultiLoRa, C-Poly, and MoE, on the GLUE, SuperGLUE, Massive Multitask Language Understanding, and commonsense reasoning benchmarks. The evaluation results present an average accuracy improvement of up to 6.67%, with individual tasks showing peak gains of up to 10.75%.

Achieving robust perception-reasoning synergy is a central goal for advanced Vision-Language Models (VLMs). Recent advancements have pursued this goal via architectural designs or agentic workflows. However, these approaches are often limited by static textual reasoning or complicated by the significant compute and engineering burden of external agentic complexity. Worse, this heavy investment does not yield proportional gains, often witnessing a "seesaw effect" on perception and reasoning. This motivates a fundamental rethinking of the true bottleneck. In this paper, we argue that the root cause of this trade-off is an ambiguity in modality credit assignment: when a VLM fails, is it due to flawed perception ("bad seeing") or flawed logic ("bad thinking")? To resolve this, we introduce a reinforcement learning framework that improves perception-reasoning synergy by reliably rewarding the perception fidelity. We explicitly decompose the generation process into interleaved perception and reasoning steps. This decoupling enables targeted supervision on perception. Crucially, we introduce Perception Verification (PV), leveraging a "blindfolded reasoning" proxy to reward perceptual fidelity independently of reasoning outcomes. Furthermore, to scale training across free-form VL tasks, we propose Structured Verbal Verification, which replaces high-variance LLM judging with structured algorithmic execution. These techniques are integrated into a Modality-Aware Credit Assignment (MoCA) mechanism, which routes rewards to the specific source of error -- either bad seeing or bad thinking -- enabling a single VLM to achieve simultaneous performance gains across a wide task spectrum.

The application of Large Language Models to Question Answering has shown great promise, but important challenges such as hallucinations and erroneous reasoning arise when using these models, particularly in knowledge-intensive, domain-specific tasks. To address these issues, we introduce Derivation Prompting, a novel prompting technique for the generation step of the Retrieval-Augmented Generation framework. Inspired by logic derivations, this method involves deriving conclusions from initial hypotheses through the systematic application of predefined rules. It constructs a derivation tree that is interpretable and adds control over the generation process. We applied this method in a specific case study, significantly reducing unacceptable answers compared to traditional RAG and long-context window methods.

Industrial LLM agent systems often separate planning from execution, yet LLM planners frequently produce structurally invalid or unnecessarily long workflows, leading to brittle failures and avoidable tool and API cost. We propose \texttt{SPIN}, a planning wrapper that combines validated Directed Acyclic Graph (DAG) planning with prefix based execution control. \texttt{SPIN} enforces a strict DAG contract through \texttt{\_validate\_plan\_text} and repair prompting, producing executable plans before downstream execution, and then evaluates DAG prefixes incrementally to stop when the current prefix is sufficient to answer the query. On AssetOpsBench, across 261 scenarios, \texttt{SPIN} reduces executed tasks from 1061 to 623 and improves \emph{Accomplished} from 0.638 to 0.706, while reducing tool calls from 11.81 to 6.82 per run. On MCP Bench, the same wrapper improves planning, grounding, and dependency related scores for both GPT OSS1 and Llama 4 Maverick.

The growing adoption of large language models in legal practice brings both significant promise and serious risk. Legal professionals stand to benefit from AI that can reason over contracts, draft documents, and analyze sources at scale, yet the high-stakes nature of legal work demands a level of rigor that current AI systems do not provide. The central problem is not simply that LLMs hallucinate facts and references; it is that they systematically draw inferences that go beyond what the source text actually supports, presenting assumption-laden conclusions as if they were logically grounded. This proposal presents a neuro-symbolic approach to legal AI that combines the expressive power of large language models with the rigor of formal verification, aiming to make AI-assisted legal reasoning both capable and trustworthy, thus reducing the burden of manual verification without sacrificing the accountability that legal practice demands.

Masked autoencoders (MAEs) have recently shown promise for self-supervised representation learning of resting-state brain functional connectivity (FC). However, a fundamental question remains unresolved: how should FC matrices be tokenized to align with the intrinsic modular organization of large-scale brain networks? Existing approaches typically adopt region-centric or graph-based schemes that treat FC as structurally homogeneous elements and overlook the large-scale network brain organization. We introduce NERVE (Network-Aware Representations of Brain Functional Connectivity via Bilinear Tokenization), a self-supervised learning framework that redefines FC tokenization by partitioning FC matrices into patches of intra- and inter-network connectivity blocks. Unlike image-based MAE, where fixed-size patches share a common tokenizer, FC patches defined by network pairs are heterogeneous in size and correspond to distinct functional roles. To resolve this problem, NERVE embeds FC patches through a novel structured bilinear factorization. This formulation preserves network identity and reduces parameter complexity from quadratic to linear scaling in the number of networks. We evaluate NERVE across three large-scale developmental cohorts (ABCD, PNC, and CCNP) for behavior and psychopathology prediction. Compared to structurally agnostic MAE variants and graph-based self-supervised baselines, the proposed network-aware formulation yields more stable and transferable representations, particularly in cross-cohort evaluation. Ablation studies confirm that the proposed bilinear network embedding and anatomically grounded parcellation are critical for performance. These findings highlight the importance of incorporating domain-specific structural priors into self-supervised learning for functional connectomics.

Deep learning excels at prediction but often lacks finite-sample guarantees and calibrated uncertainty; RKHS (Reproducing Kernel Hilbert Space)-based methods provide those guarantees but struggle to adapt in high dimensions. We propose Wahkon, a deep RKHS superposition network that unifies Kolmogorov's superposition principle with RKHS regularization in the smoothing-spline tradition of Wahba. This yields a finite-dimensional deep representer theorem that makes training tractable and provides explicit layerwise complexity control. We show the penalized estimator is exactly the MAP (maximum a posteriori) estimate under a hierarchical Gaussian-process prior, extending the spline/GP duality to deep compositions. Using metric-entropy arguments, we establish minimax-optimal convergence rates under mild smoothness and clarify how depth and width trade off with regularity. Empirically, Wahkon outperforms multilayer perceptrons, Neural Tangent Kernels, and Kolmogorov--Arnold Networks across simulation benchmarks and a single-cell CITE-seq study. By unifying Kolmogorov's superposition principle with RKHS regularization, Wahkon delivers accuracy, interpretability, and statistical rigor in a single framework.

We audit the multimodal-physics evaluation pipeline end-to-end and document three undetected construction practices that distort how the field measures vision-language reasoning: train-eval contamination, translation drift, and MCQ saturation. (1) Public training pools (UGPhysics-Train, SciInstruct, MMK12) pass single-stage 5-gram-Jaccard audits with zero hits across all six public physics evals; a three-stage audit (Jaccard -> mxbai-embed-large cosine -> Haiku-4.5 LLM-judge) surfaces 134 near-duplicates and 4,846 paraphrase candidates in SciInstruct alone. (2) A 17-pp Sonnet 4.5 delta on 59 paired Estonian-English olympiad problems (30.5% vs. 13.6%; sign test p=0.011, McNemar p=0.021, paired bootstrap 95% CI [+5.1, +28.9] pp). (3) A 46-pp format-and-novelty gradient on identical Sonnet weights between MCQ (79.7% on PhyX) and open-ended olympiad evaluation (33.4% on PhysOlym-A). We release four artifacts addressing these gaps: PhysCorp-A (6,432-record three-stage-audited multimodal corpus), PhysR1Corp (2,268-record closed-form RL pool), PhysOlym-A (500-problem, 99.8% novel-source held-out olympiad eval with native difficulty labels and an EN/ET bilingual subset), and Physics-R1, a reference GSPO+DAPO recipe cold-started from Qwen3-VL-8B-Thinking. Across 3 seeds, Physics-R1 lifts the audited corpus over the 8B base by +18.3 pp on PhysOlym-A liberal (8.0 -> 26.3 +/- 1.7; 7.1 pp behind Sonnet 4.5), +15.7 pp on PhysReason (23.9 -> 39.6 +/- 6.4; ahead of Qwen3-VL-32B and Gemini 2.5 Pro), +6.9 pp on OlympiadBench-Physics (46.2 +/- 1.5), and +4.1 pp on PhyX MCQ (77.8 +/- 0.3).

Large language models (LLMs) increasingly act as autonomous agents that must decide when to answer directly vs. when to invoke external tools. Prior work studying adaptive tool use has largely treated tool necessity as a model-agnostic property, annotated by human or LLM judge, and mostly cover cases where the answer is obvious (e.g., fetching the weather vs. paraphrasing text). However, tool necessity in the wild is more nuanced due to the divergence of capability boundaries across models: a problem solvable by a strong model on its own may still require tools for a weaker one. In this work, we introduce a model-adaptive definition of tool-necessity, grounded in each model's empirical performance. Following this definition, we compare the necessity against observed tool-call behavior across four models on arithmetic and factual QA dataset, and find substantial mismatches of 26.5-54.0% and 30.8-41.8%, respectively. To diagnose the failure, we decompose tool use into two stages: an internal cognition stage that reflects whether a model believes a tool is necessary, and an execution stage that determines whether the model actually makes a tool-call action. By probing the LLM hidden states, we find that both signals are often linearly decodable, yet their probe directions become nearly orthogonal in the late-layer, last-token regime that drives the next-token action. By tracing the trajectory of samples in the two-stage process, we further discover that the majority of mismatch is concentrated in the cognition-to-action transition, not in cognition itself. These results reveal a knowing-doing gap in LLM tool-use: improving tool-use reliability requires not only better recognition of when tools are needed, but also better translation of that recognition into action.

Under modern test-time compute and agentic paradigms, language models process ever-longer sequences. Efficient text generation with transformer architectures is increasingly constrained by the Key-Value cache memory footprint and bandwidth. To address this limitation, we introduce Self-Pruned Key-Value Attention (SP-KV), a mechanism designed to predict future KV utility in order to reduce the size of the long-term KV cache. This strategy operates at a fine granularity: a lightweight utility predictor scores each key-value pair, and while recent KVs are always available via a local window, older pairs are written in the cache and used in global attention only if their predicted utility surpasses a given threshold. The LLM and the utility predictor are trained jointly end-to-end exclusively through next-token prediction loss, and are adapted from pretrained LLM checkpoints. Rather than enforcing a fixed compression ratio, SP-KV performs dynamic sparsification: the mechanism adapts to the input and typically reduces the KV cache size by a factor of $3$ to $10\times$, longer sequences often being more compressible. This leads to vast improvements in memory usage and decoding speed, with little to no degradation of validation loss nor performance on a broad set of downstream tasks. Beyond serving as an effective KV-cache reduction mechanism, our method reveals structured layer- and head-specific sparsity patterns that we can use to guide the design of hybrid local-global attention architectures.

In current Large Language Models we can trust the production of smoothly flowing prose on the basis of the principles of machine learning. However, there is no comparably principled basis to justify trust in the content of the text produced. It appears to be conventional wisdom that addressing this issue by adding more principled reasoning is not computationally affordable. Here we propose a principled method of reasoning that is efficient enough to be practical for large language models. Further, the method allows the retention of much of the currently used software and hardware base. Our method for improving the functioning of large language models consists of a first stage of preprocessing that recodes the data to a Unary Relational Integracode that is more explicit about the relationships among the objects described in the text, followed as a second stage by a standard but possibly streamlined machine learning process that then also learns to predict these relationships. The method may be viewed as realizing a world model and applying beyond natural language, to vision and actions, for example, where the multiple properties of an object referred to in an input are brought together explicitly, rather than remaining distributed in the various references to it in the input. We articulate its advantages in terms of Robust Logic, a system for performing principled chaining on learned, and hence uncertain, information. We show that this recoding has the surprising and fortuitous property that, while succinct, it makes the task of learning a core subset of relational rules that hold in the world described in the training data polynomial time learnable in a defined sense, the polynomial depending on the complexity of the rule. This gives support for sound reasoning within each single call of the learned classifier as well as between multiple calls.

Wide applications of LLM-based agents require strong alignment with human social values. However, current works still exhibit deficiencies in self-cognition and dilemma decision, as well as self-emotions. To remedy this, we propose a novel value-based framework that employs GraphRAG to convert principles into value-based instructions and steer the agent to behave as expected by retrieving the suitable instruction upon a specific conversation context. To evaluate the ratio of expected behaviors, we define the expected behaviors from two famous theories, Maslow's Hierarchy of Needs and Plutchik's Wheel of Emotion. By experimenting with our method on the benchmark of DAILYDILEMMAS, our method exhibits significant performance gains compared to prompt-based baselines, including ECoT, Plan-and-Solve, and Metacognitive prompting. Our method provides a basis for the emergence of self-emotion in AI systems.

Scientific theory shift in AI agents requires more than fitting equations to data. An artificial scientific agent must detect whether an existing representational framework remains transportable into a new regime, or whether its language has become locally-to-globally obstructed and must be extended. This paper develops a finite sheaf-theoretic framework for detecting theory-shift candidates through transport and obstruction. Contexts are organized as a local-to-global structure in which source, overlap, target, and validation charts are fitted, restricted, and tested for gluing. Obstruction measures failure of coherence through residual fit, overlap incompatibility, constraint violation, limiting-relation failure, and representational cost. We evaluate the framework on a controlled transition-card benchmark designed to separate deformation within a source language from extension of that language. The main result is direct obstruction ranking: the intended deformation or extension is usually the lowest-obstruction candidate, and transition type is separated in the benchmark. A constellation kernel over the same signatures is included only as a secondary representational-similarity probe. The aim is not to reconstruct historical paradigm shifts or solve open-ended autonomous theory invention, but to isolate a finite diagnostic subproblem for AI agents: detecting when representational transport fails and extension becomes the coherent next move.

Bioacoustic recognition requires fine-grained acoustic understanding to distinguish similar-sounding species. However, many large-scale data repositories such as iNaturalist are weakly annotated, often with only a single positive species label per recording, making supervised learning particularly challenging. Inspired by advances in computer vision, recent approaches have shifted toward self-supervised learning to capture the underlying structure of audio without relying on exhaustive annotations. In particular, masked autoencoders (MAE) have shown strong transferability on massive audio corpora, yet their effectiveness in more modest bioacoustic settings remains underexplored. In this work, we conduct a systematic study of MAE pretraining for species classification on iNatSounds, analyzing the impacts of pretraining data scale, domain specificity, data curation, and transfer strategies. Consistent with prior work, we find that models pretrained on diverse general audio data achieve the best transfer performance on iNatSounds. Contrary to observations from large-scale audio benchmarks, we find that (1) additional masked reconstruction pretraining on domain-specific data provides limited benefits and may even degrade performance relative to off-the-shelf models, and (2) selective data filtering offers a negligible advantage when the overall data scale is limited. Our results indicate that, in moderate-sized fine-grained bioacoustic settings, pretraining scale dominates objective design. These findings further clarify when MAE-based pretraining is effective and provide practical guidance for model selection under limited supervision.

For reinforcement learning in data-scarce domains like real-world robotics, intensive data reuse enhances efficiency but induces overfitting. While prior works focus on critic bias, representation-level instability in Self-Predictive Learning (SPL) under high Update-to-Data (UTD) regimes remains underexplored. To bridge this gap, we propose Robust Representation via Redundancy Reduction (R2R2), a regularization method within SPL. We theoretically identify that standard zero-centering conflicts with SPL's spectral properties and design a non-centered objective accordingly. We verify R2R2 on SPL-native algorithms like TD7. Furthermore, to demonstrate its orthogonality to prior advancements, we extend the state-of-the-art SimbaV2, which originally lacks SPL, by integrating a tailored SPL module, termed SimbaV2-SPL. Experiments across 11 continuous control tasks confirm that R2R2 effectively mitigates overfitting; specifically, at a UTD ratio of 20, it improves TD7 by ~22% and provides additional gains on top of SimbaV2-SPL, which itself establishes a new state-of-the-art. The code can be found at: https://github.com/songsang7/R2R2

Brain-language model comparisons often interpret neural prediction scores as evidence that model representations capture brain-relevant language computation. We asked whether language models align with brains, and whether prediction scores are enough to support that claim, using L-PACT, a source-audited framework that evaluates predictive, relational, mechanism-stripping, and reliability-bounded evidence. Across primary naturalistic language neural datasets and derived language-model representations, L-PACT compared real model features with nuisance baselines and severe controls, tested whether model-to-brain profiles reproduced brain-to-brain patterns, recomputed held-out scores after mechanism stripping, and normalized evidence against brain-brain ceilings. The locked analysis set contains 414 predictive-control rows, 2304 relational profile rows, 4320 mechanism-stripping rows, 420 brain-brain ceiling rows, and 146 integrated decision rows. Assay-sensitivity checks showed that brain-brain reliability, brain-as-model run-to-run relational profiles, independent low-level neural and WAV-derived acoustic-envelope gates, and a deterministic implanted-signal simulation can produce positive evidence when expected. Nevertheless, no real model row passed the predictive, relational, mechanism-stripping, or operational Turing-bounded reliability gates; all 146 integrated rows were control-explained. Less stringent single-criterion rules would have counted raw positive predictive, relational, stripping-delta, and ceiling-normalized effects, but L-PACT downgraded them because controls explained the apparent evidence. In the analyzed derived artifact set, the tested language-model representations do not satisfy L-PACT alignment gates; apparent positives are converted into an auditable control-explained taxonomy rather than treated as structural alignment.

Google AI Overviews (AIOs) are arguably the most widely encountered deployment of generative AI, reaching over 2 billion users who may not realize the answers they see are AI-generated. Where search engines have traditionally surfaced ranked sources and left users to evaluate them, AIOs synthesize and deliver a single answer - giving Google unprecedented editorial control over what users read and know. We present a large-scale longitudinal measurement study, issuing 55,393 trending queries across 19 topical categories over a 40-day window (March 13 - April 21, 2026). We report four main findings. First, overall AIO activation is 13.7%, rising to 64.7% for question-form queries, while politically sensitive topics see markedly lower rates. Second, AIO-cited domains are more credible than co-displayed first-page results, yet nearly 30% do not appear in those results at all, indicating a source selection mechanism distinct from Google's ranking algorithm. Third, decomposing responses into 98,020 atomic claims, 11.0% are unsupported by the cited pages - with omission the dominant failure mode - and source quality and claim fidelity are largely independent. Fourth, well over half of AIO-cited pages carry display advertising, meaning publishers lose revenue when AIOs suppress the click-through, even as Google's own sponsored ads continue to appear on the same page. Together, these findings document a rapid transformation of the online information ecosystem whose consequences for epistemic security remain poorly understood.

Regret is the cost of uncertainty in algorithmic decision-making. Quantifying regret typically requires computationally expensive simulation via Sample Average Approximation (SAA), with complexity $\mathcal{O}(Bn^{2}d^{3})$ in the number of scenarios $B$, variables $n$, and constraints $d$. % This paper proves that expected regret in any stochastic optimization problem admits the exact decomposition % \begin{equation*} \mathrm{Regret}(c) = \mathrm{Cov}(c,\,π^{*}(c)) + R(c), \end{equation*} % where $c$ is the vector of uncertain parameters, $π^{*}(c)$ is the optimal decision, and $R(c)$ is a residual whose magnitude we bound explicitly under Lipschitz, smooth, and strongly convex conditions. % For linear programs and unconstrained quadratic programs, including the classical Markowitz portfolio problem, we prove $R(c)=0$ exactly, so that $\mathrm{Regret}(c) = \mathrm{Cov}(c,π^{*}(c))$ holds without approximation. % When historical cost-decision pairs $\{(c_i, π^*(c_i))\}$ are available, the covariance can be estimated in $\mathcal{O}(nd^{2})$ time, which is orders of magnitude faster than SAA. The estimation is performed by a single pass through the data. % We derive concentration bounds, a central limit theorem, and an asymptotically unbiased residual estimator, and we validate all results on synthetic LP, QP, and integer programming instances and on a rolling-window portfolio experiment using ten years of CRSP equity data.

The article explores the use of agentic software engineering (ASE) in the development of innovative audio software. It begins with a review of background work that lays out the challenges of longevity, interoperability and barriers to entry in digital music instrument creation, explaining recent developments in ASE and highlighting the possibility that ASE can lower barriers to entry and facilitate creation of interoperable software with greater longevity. Following that, we present case studies wherein we used ASE technology in three distinct ways to develop audio software in the C++ language with the JUCE framework. In case study 1, we re-implement Laurie Spiegel's `Music Mouse' software as a native plugin. In case study 2, we translate Pachet's `Continuator' system from Python into a native plugin. In case study 3, we develop a new 3D user interface for an existing `tracker' sequencer using OpenGL. We describe the experiences of the human developer in the case studies via autoethnographic discussion of the prompt logs and snapshots of the software as it was developed. We identify effective practice for ASE use in this domain and suggest future steps for the work involving evaluation of the method with non-programmer musicians.

Internet of Things (IoT) systems continuously collect heterogeneous sensing signals from ubiquitous sensors to support intelligent applications such as human activity analysis, emotion monitoring, and environmental perception. These signals are inherently non-stationary and multi-scale, posing unique challenges for standard tokenization techniques. This paper proposes Dywave, a dynamic tokenization framework for IoT sensing signals that constructs compact input representations aligned with intrinsic temporal structures and underlying physical events. Dywave leverages wavelet-based hierarchical decomposition, identifies meaningful temporal boundaries corresponding to underlying semantic events, and adaptively compresses redundant intervals while preserving temporal coherence. Extensive evaluations on five real-world IoT sensing datasets across activity recognition, stress assessment, and nearby object detection demonstrate that Dywave outperforms state-of-the-art methods by up to 12% in accuracy, while improving computational efficiency by reducing input token lengths by up to 75% across mainstream sequence models. Moreover, Dywave exhibits improved robustness to domain shifts and varying sequence lengths.

Speculative decoding has become a widely adopted technique for accelerating large language model (LLM) inference by drafting multiple candidate tokens and verifying them with a target model in parallel. Its efficiency, however, critically depends on the average accepted length $τ$, i.e., how many draft tokens survive each verification step. In this work, we identify a new mechanism-level vulnerability in model-based speculative decoding: the drafter is trained to approximate the target model distribution, but this approximation is inevitably imperfect. Such a drafter-target mismatch creates a hidden attack surface where small perturbations can preserve the target model's visible behavior while substantially reducing draft-token acceptability. We propose Mistletoe, a stealthy acceleration-collapse attack against speculative decoding. Mistletoe directly targets the acceptance mechanism of speculative decoding. It jointly optimizes a degradation objective that decreases drafter-target agreement and a semantic-preservation objective that constrains the target model's output distribution. To resolve the conflict between these objectives, we introduce a null-space projection mechanism, where degradation gradients are projected away from the local semantic-preserving direction, suppressing draft acceptance while minimizing semantic drift. Experiments on various speculative decoding systems show that Mistletoe substantially reduces average accepted length $τ$, collapses speedup, and lowers averaged token throughput, while preserving output quality and perplexity. Our work highlights that speculative decoding introduces a mechanism-level attack surface beyond existing output robustness, calling for more robust designs of LLM acceleration systems.

Generative models are often trained with a next-token prediction objective, yet many downstream applications require the ability to estimate or control sequence-level properties. Next-token prediction can lead to overfitting of local patterns during training, underfitting of global structure, and requires significant downstream modifications or expensive sampling to guide or predict the global attributes of generated samples at inference time. Here, we introduce Conditional Attribute Transformers, a novel method for jointly estimating the next-token probability and the value of an attribute conditional on each potential next token selection. This framework enables three critical capabilities within a single forward pass, without modification of the input sequence: (1) per-token credit assignment across an entire sequence, by identifying how each token in a sequence is associated with an attribute's value; (2) counterfactual analysis, by quantifying attribute differences conditional on alternative next token choices; (3) steerable generation, by decoding sequences based on a combination of next-token and attribute likelihoods. Our approach achieves state of the art performance on sparse reward tasks, improves next-token prediction at sufficient model sizes, estimates attribute probabilities orders of magnitude faster than sampling, and can guide decoding of autoregressive sequence models on a range of language tasks.

Large Reasoning Models (LRMs) embedded in agentic frameworks have transformed information retrieval from static, long context question answering into open-ended exploration. Yet real world use requires models to discover and synthesize "long-tail" facts from dispersed sources, a capability that remains under-evaluated. We introduce PolitNuggets, a multilingual benchmark for agentic information synthesis via constructing political biographies for 400 global elites, covering over 10000 political facts. We standardize evaluation with an optimized multi agent system and propose FactNet, an evidence conditional protocol that scores discovery, fine-grained accuracy, and efficiency. Across models and settings, we find that current systems often struggle with fine-grained details, and vary substantially in efficiency. Finally, using benchmark diagnostics, we relate agent performance to underlying model capabilities, highlighting the importance of short-context extraction, multilingual robustness, and reliable tool use.

Discrete diffusion models are increasingly competitive for language modeling, yet it remains unclear how their denoising objectives organize learning. Although these objectives target the full data distribution, we show that the exact reverse process induces a hierarchy between coarse support information and finer frequency information. For uniform and absorbing (a.k.a. masking) diffusion, we prove that, in the small-noise regime of the final denoising steps, each single-token reverse edit decomposes into a leading scale, determined by whether it moves toward the data support (e.g., grammatically valid sentences), and a finer coefficient, determining relative probabilities within the same scale. Thus, recovering validity structure only requires learning the correct order of magnitude of reverse probabilities, whereas recovering data frequencies requires coefficient-level estimation. The separation is mechanism-dependent: uniform diffusion exhibits a trichotomy into validity-improving, validity-preserving, and validity-worsening edits, while absorbing diffusion places its leading-order mass on validity-improving moves. Experiments on a masked language diffusion model and synthetic regular-language tasks support these predictions: support-localization emerges earlier than within-support frequency ranking, and the contrast between uniform and absorbing diffusion matches the predicted rate separation. Together, our results suggest that discrete diffusion models learn data support before data frequencies.

Generating realistic synthetic option prices requires implied volatility as an input, yet implied volatility is itself derived from observed option prices, creating a circular dependency that limits synthetic data for machine-learning and risk-analysis applications. We break this circularity with a pipeline in which implied volatility emerges as an output of a structural model of equity returns. A Jump Hidden Markov Model produces multi-asset price paths with realistic stylized facts and cross-asset tail dependence; a modified Heston variance process, whose mean-reversion target depends on regime state, days to expiration, moneyness, and a market-mood indicator, converts those paths into implied-volatility paths; and a recombining binomial lattice prices American options from the resulting surface. Initializing variance at its mean-reversion target for each strike-expiration pair lets smile, skew, and term structure emerge without external calibration. We calibrate the shape function through a hierarchy spanning a parametric baseline, a globally shared neural surrogate, and a sector-specific neural surrogate fit to a multi-ticker, multi-sector option ladder. A temporal holdout on a multi-day capture isolated scheduled corporate events as the dominant source of test-time generalization error, and calendar-derived earnings-distance and same-sector peer-coupling features recovered the anticipatory portion of that signal. We then apply the framework as a synthetic-data generator on real near-the-money put and call contracts, forward-simulating price paths, and recovering path-conditional implied volatility, finite-difference American Greeks, and terminal short-premium profit and loss from one coherent simulation, and confirm cross-ticker robustness by re-running on a second underlying from a different sector and volatility regime. The framework is released as an open-source Julia package.

Sparse Mixture-of-Experts (MoE) layers route tokens through a handful of experts, and learning-free compression of these layers reduces inference cost without retraining. A subtle obstruction blocks every existing compressor in this family: three experts can each be pairwise compatible yet form an irreducible cycle when merged together, so any score that ranks experts on pairwise signals is structurally blind to which triples are jointly mergeable. We show the obstruction is a precise mathematical object, the harmonic kernel of the simplicial Laplacian on a 2-complex whose vertices are experts, whose edges carry KL merge barriers, and whose faces carry triplet barriers; Hodge-decomposing the edge-barrier signal isolates the kernel exactly. We turn the diagnostic into a selection objective: HodgeCover greedily covers the harmonic-critical edges and triplet-critical triangles, and a hybrid variant of HodgeCover pairs it with off-the-shelf weight pruning on survivors. On three open-weight Sparse MoE backbones under aggressive expert reduction, HodgeCover matches state-of-the-art learning-free baselines on the expert-reduction axis, leads on the aggressive-compression frontier of the hybrid axis, and uniquely balances retained mass across all four Hodge components. These results show that exposing the harmonic kernel of a learned MoE structure changes which compressor wins at the regime that matters most.

Accurate 3D+t whole-heart mesh reconstruction from cine MRI is a clinically crucial yet technically challenging task. The difficulty of this task arises from two coupled factors: inherently sparse sampling of 3D cardiac anatomy by 2D image slices and the tight coupling between cardiac shape and motion. Current cardiac image-to-mesh approaches typically reconstruct only a subset of cardiac chambers or a single phase of the cardiac cycle. In this work, we propose CineMesh4D, a novel end-to-end 4D (3D+t) pipeline that directly reconstructs patient-specific whole-heart mesh from multi-view 2D cine MRI via cross-domain mapping. Specifically, we introduce a differentiable rendering loss that enables supervision of 3D+t whole-heart mesh from multi-view sparse contours of cine MRI. Furthermore, we develop a dual-context temporal block that fuses global and local cardiac temporal information to capture high-dimensional sequential patterns. In quantitative and qualitative evaluations, CineMesh4D outperforms existing approaches in terms of reconstruction quality and motion consistency, providing a practical pathway for personalized real-time cardiac assessment. The code will be publicly released once the manuscript is accepted.

We present VectraYX-Nano, a 41.95M-parameter decoder-only language model trained from scratch in Spanish for cybersecurity, with a Latin-American focus and native tool invocation via the Model Context Protocol (MCP). Four contributions: (i) Corpus: VectraYX-Sec-ES, a 170M-token Spanish corpus from an eight-VM pipeline (~$25 USD) partitioned into conversational (42M tokens, OpenSubtitles-ES, OASST1), cybersecurity (118M tokens, NVD, Wikipedia-ES, CVE mirror, security blogs), and offensive-security tooling (10M tokens, ExploitDB, HackTricks, OWASP) phases. (ii) Architecture: 42M-parameter Transformer decoder with GQA, QK-Norm, RMSNorm, SwiGLU, RoPE, z-loss, and a 16,384-token byte-fallback BPE. (iii) Curriculum with replay: continual pre-training with a replay buffer yields monotonic loss descent (9.80->3.17->3.00->2.16); after SFT on OASST-ES, Alpaca-ES, CVE Q&A, and 6,327 tool-use traces, the model attains a conversational gate of 0.78+-0.05 (N=4 seeds). (iv) Two findings: a bootstrap-corpus ablation reveals a loss-vs-register inversion at nano scale; a LoRA study shows the B4 tool-selection floor of 0.000 is a corpus-density artifact, not a capacity gate -- a tool-dense corpus (2,801 examples) raises B4 to 0.145+-0.046 on Nano 42M and 0.445+-0.201 on a 260M mid-tier. The GGUF artifact is 81 MB (F16), runs at sub-second TTFT on commodity hardware under llama.cpp, and is to our knowledge the first Spanish-native cybersecurity LLM with end-to-end MCP integration. Corpus recipe, training scripts, GGUF weights, and B1-B5 benchmark are released.

Inverse problems in scientific sensing are often solved with either hand-designed regularizers or supervised networks trained on simulated labels, yet both can fail when the forward model is nonlinear, spectrally coupled, and physically delicate. We study this issue for noise sensing based on nitrogen-vacancy (NV) centers in diamond, where a quantum sensor measures magnetic-noise spectra generated by sparse spin sources. We show that replacing a common scalar/coherent forward approximation with a tensor power-summed dipolar operator changes the inverse landscape and exposes a center-collapse failure mode in free-density optimization. We propose NeTMY, an amortization-free coordinate neural field coupled to the differentiable NV forward model, with annealed positional encoding, multiscale optimization, sparsity/gating, and spectrum-fidelity losses. Across sparse synthetic reconstructions generated by the corrected operator, NeTMY achieves the best localization and distributional metrics in the tested benchmark. Mechanism experiments show that NeTMY does not directly execute the raw density-space gradient; its parameterization smooths and redistributes updates, mitigating the center-collapse pathology. These results position NV quantum sensing as a useful testbed for physics-faithful neural inverse problems.

Tabular data underpins most high-value prediction problems in science and industry, and TabPFN has driven the foundation model revolution for this modality. Designed with feedback from our users, TabPFN-3 builds on this foundation to scale state-of-the-art performance to datasets with 1M training rows and substantially reduce training and inference time. Pretrained exclusively on synthetic data from our prior, TabPFN-3 dramatically pushes the frontier of tabular prediction and brings substantial gains on time series, relational, and tabular-text data. On the standard tabular benchmark TabArena, a forward pass of TabPFN-3 outperforms all other models, including tuned and ensembled baselines, by a significant margin, and pareto-dominates the speed/performance frontier. On more diverse datasets, TabPFN-3 ranks first on datasets with many classes, and beats 8-hour-tuned gradient-boosted-tree baselines on datasets up to 1M training rows and 200 features. TabPFN-3 introduces test-time compute scaling to tabular foundation models. Our API offering TabPFN-3-Plus (Thinking) exploits this to beat all non-TabPFN models by over 200 Elo on TabArena, rising to 420 Elo on the largest data subset, and outperforms AutoGluon 1.5 extreme while being 10x faster, without using LLMs, real data, internet search or any other model besides TabPFN. TabPFN-3 extends the capabilities of our models, enabling SOTA prediction on relational data (new SOTA foundation model on RelBenchV1) and tabular-text data (SOTA on TabSTAR via TabPFN-3-Plus); and improves existing integrations: a specialized checkpoint, TabPFN-TS-3, ranks 2nd on the time-series benchmark fev-bench, and SHAP-value computation is up to 120x faster. TabPFN-3 achieves this performance while being up to 20x faster than TabPFN-2.5. In addition, a reduced KV cache and row-chunking scale to 1M rows on one H100 with fast inference speed.

The rise in deployment of large language models has driven a surge in GPU demand and datacenter scaling, raising concerns about electricity use, grid stress, and the impacts of modern AI workloads. Distillation is often promoted as one of the most effective paths to obtain cheaper, more efficient models, yet these claims rarely account for the full end-to-end energy and resource costs, including crucial teacher-side workloads such as data generation, logit caching, and evaluation. We present a comprehensive energy accounting framework that measures the complete computational cost of distillation pipelines via detailed stage-wise tracking of GPU device power consumption. In our experiments, we separate and log empirical energy use across distinct phases and systematically measure the energy and emissions of two common distillation methods: the classic logit-based knowledge distillation and synthetic-data supervised fine-tuning, constructing energy-quality Pareto frontiers that expose the previously ignored costs. From these measurements and analyses, we derive practical design rules for selecting distillation methods and hyperparameters under energy and budget constraints, and release an open-source measurement harness and accounting protocol to provide a standardized foundation for comparable, reproducible distillation research, explicitly accountable for complete pipeline energy impact.

Quantum machine learning (QML) aims to accelerate machine learning tasks by exploiting quantum computation. Previous work studied a QML algorithm for selecting sparse subnetworks from large shallow neural networks. Instead of directly solving an optimization problem over a large-scale network, this algorithm constructs a sparse subnetwork by sampling hidden nodes from an optimized probability distribution defined using the ridgelet transform. The quantum algorithm performs this sampling in time $O(D)$ in the data dimension $D$, whereas a naive classical implementation relies on handling exponentially many candidate nodes and hence takes $\exp[O(D)]$ time. In this work, we construct and analyze a quantum-inspired fully classical algorithm for the same sampling task. We show that our algorithm runs in time $O(\operatorname{poly}(D))$, thereby removing the exponential dependence on $D$ from the previous classical approach. Numerical simulations show that the proposed sampler achieves empirical risk comparable to exact sampling from the optimized distribution and substantially lower than sampling from the non-optimized uniform distribution, while also exhibiting exponentially improved runtime scaling compared with the conventional classical implementation. These successful dequantization results show that sparse subnetwork selection via optimized sampling can be achieved classically with polynomial data-dimension scaling on conventional computers without quantum hardware, providing an alternative to the existing quantum algorithm.

Diffusion Transformers (DiTs) and related flow-based architectures are now among the strongest text-to-image generators, yet the internal mechanisms through which prompts shape image semantics remain poorly understood. In this work, we study massive activations: a small subset of hidden-state channels whose responses are consistently much larger than the rest. We show that, despite their sparsity, these few channels effectively draw the whole picture, in three complementary senses. First, they are functionally critical: a controlled disruption probe that zeroes the massive channels causes a sharp collapse in generation quality, while disrupting an equally-sized set of low-statistic channels has marginal effect. Second, they are spatially organized: restricting image-stream tokens to massive channels and clustering them yields coherent partitions that closely align with the main subject and salient regions, exposing a structured spatial code hidden inside an apparently outlier-like subspace. Third, they are transferable: transporting massive activations from one prompt-conditioned trajectory into another, shifts the final image toward the source prompt while preserving substantial content from the target, producing localized semantic interpolation rather than unstructured pixel blending. We exploit this property in two use cases: text-conditioned and image-conditioned semantic transport, where massive activations transport enables prompt interpolation and subject-driven generation without any additional training. Together, these results recast massive activations not as activation anomalies, but as a sparse prompt-conditioned carrier subspace that organizes and controls semantic information in modern DiT models.

Generative policies based on diffusion and flow matching have become a dominant paradigm for visuomotor robotic control. We show that replacing the standard Gaussian source distribution with WarmPrior, a simple temporally grounded prior constructed from readily available recent action history, consistently improves success rates on robotic manipulation tasks. We trace this gain to markedly straighter probability paths, echoing the effect of optimal-transport couplings in Rectified Flow. Beyond standard behavior cloning, WarmPrior also reshapes the exploration distribution in prior-space reinforcement learning, improving both sample efficiency and final performance. Collectively, these results identify the source distribution as an important and underexplored design axis in generative robot control.

Autonomous language-model agents are increasingly evaluated on long-horizon tool-use tasks, but existing benchmarks rarely capture the complexity and nuance of real scientific work. To address this gap, we introduce Collider-Bench, a benchmark for evaluating whether LLM agents can reproduce experimental analyses from the Large Hadron Collider (LHC) using only public papers and open scientific software. Such analyses are often difficult to reproduce because the public toolchain only approximates the software used internally by the experimental collaborations, while the published papers inevitably omit implementation details needed for a faithful reconstruction. Agents must therefore rely on physical reasoning, domain knowledge, and trial-and-error to fill these gaps. Each task requires the agent to turn a published analysis into an executable simulation-and-selection pipeline and submit predicted collision event yields in specified signal regions. These predictions are evaluated with standard histogram metrics that provide continuous fidelity scores without a hand-written rubric. We also report the computational cost incurred by each agent per task. Finally, we evaluate the codebase and full session trace using an LLM judge to catch qualitative failure modes such as fabrications, hallucinations and duplications. We release an initial set of tasks drawn from LHC searches, together with a containerized sandbox and event simulation tools. We evaluate across a capability ladder of general purpose coding agents. Our results show that on average no agent reliably beats the physicist-in-the-loop solution.

This paper introduces WARDEN, an early language model system capable of transcribing and translating Wardaman, an endangered Australian indigenous language into English. The significant challenge we face is the lack of large-scale training data: in fact, we only have 6 hours of annotated audio. Therefore, while it is common practice to train a single model for transcription and translation using large datasets (like English to French), this practice is no longer viable in the Wardaman to English context. To tackle the low-resource challenge, we design WARDEN to have separate transcription and translation models: WARDEN first turns a Wardaman audio input into phonemic transcription, and then the transcription into English translation. Further, we propose two useful techniques to enhance performance. For transcription, we initialize the Wardaman token from Sundanese, a language that shares similar phonemes with Wardaman, to accelerate fine-tuning of the transcription model. For translation, we compile a Wardaman-English dictionary from expert annotations, and provide this domain-specific knowledge to a large language model (LLM) to reason and decide the final output. We empirically demonstrate that this two-stage design works better than data-hungry unified approaches in extremely low data settings. Using a mere 6 hours of annotated data, WARDEN outperforms larger open-source and proprietary models and establishes a strong baseline. Data and code are available.

Voice agents, artificial intelligence systems that conduct spoken conversations to complete tasks, are increasingly deployed across enterprise applications. However, no existing benchmark jointly addresses two core evaluation challenges: generating realistic simulated conversations, and measuring quality across the full scope of voice-specific failure modes. We present EVA-Bench, an end-to-end evaluation framework that addresses both. On the simulation side, EVA-Bench orchestrates bot-to-bot audio conversations over dynamic multi-turn dialogues, with automatic simulation validation that detects user simulator error and appropriately regenerates conversations before scoring. On the measurement side, EVA-Bench introduces two composite metrics: EVA-A (Accuracy), capturing task completion, faithfulness, and audio-level speech fidelity; and EVA-X (Experience), capturing conversation progression, spoken conciseness, and turn-taking timing. Both metrics apply to different agent architectures, enabling direct cross-architecture comparison. EVA-Bench includes 213 scenarios across three enterprise domains, a controlled perturbation suite for accent and noise robustness, and pass@1, pass@k, pass^k measurements that distinguish peak from reliable capability. Across 12 systems spanning all three architectures, we find: (1) no system simultaneously exceeds 0.5 on both EVA-A pass@1 and EVA-X pass@1; (2) peak and reliable performance diverge substantially (median pass@k - pass^k gap of 0.44 on EVA-A); and (3) accent and noise perturbations expose substantial robustness gaps, with effects varying across architectures, systems, and metrics (mean up to 0.314). We release the full framework, evaluation suite, and benchmark data under an open-source license.

Valiant's 1984 paper is widely credited with introducing the PAC learning model, but it, in fact, introduced a different model: unlike PAC learning, the learner receives only positives, may issue membership queries, and must output a hypothesis with no false positives. Prior work characterized variants, including the case without queries. We revisit Valiant's original model and ask: *Which classes are learnable in it?* For every finite domain, including Valiant's Boolean-hypercube setting, we show that a class is learnable if and only if every realizable positive sample can be certified by a poly-size adaptive query-compression scheme. This is a new variant of sample compression where the learner certifies samples via a short interaction with the membership oracle. Our characterization shows that learnability in Valiant's model is strictly sandwiched between learnability in the PAC model and the variant of Valiant's model without membership queries. This is one of the rare cases where introducing membership queries changes the set of learnable classes, and not just the sample or computational complexity. Next, we study the natural extension of the model to arbitrary domains. While we do not obtain an exact characterization, our techniques readily generalize and show that the same strict sandwiching persists. Finally, we show that $d$-dimensional halfspaces, which are not learnable without queries, are learnable with queries: we give a $\mathrm{poly}(d) \tilde{O}(1/ε)$ sample and $\mathrm{poly}(d) \mathrm{polylog}(1/ε)$ query algorithm, and prove that at least $Ω(d)$ samples or queries are necessary. To our knowledge, this is the first algorithm for halfspaces in Valiant's model. Together, these results uncover a surprisingly rich theory behind Valiant's original notion of learnability and introduce ideas that may be of independent interest in learning theory.

Multi-agent LLM systems usually collaborate by exchanging natural-language messages. This interface is simple and interpretable, but it forces each sender's intermediate computation to be serialized into tokens and then reprocessed by the receiver, thereby increasing the generated-token cost, prefill overhead, and KV-cache memory. We study an alternative communication interface: instead of appending a sender's message to the receiver's context, compile the sender's hidden states into a transient, receiver-specific weight perturbation. We introduce TFlow (Thought Flow), a weight-space communication framework for a known and fixed receiver architecture. For each query, frozen role-prompted sender agents process the input, and a learned parameter generator maps their internal activations into low-rank LoRA perturbations targeting the receiver's modules. These perturbations are fused and applied only during the receiver's generation, enabling instance-level adaptation without permanently changing the model or enlarging the receiver's text context. With three Qwen3-4B agents, TFlow improves over a standalone receiver by up to 8.5 accuracy points across five benchmarks while reducing processed tokens by up to 32.69%. Compared with a text-based three-agent baseline, it reduces total processed tokens by up to 83.27% and the wall-clock inference time by up to 4.6$\times$, while maintaining competitive accuracy on four of five benchmarks. These results suggest that transient low-rank weight perturbations can serve as an executable communication medium for efficient multi-agent LLM collaboration.

Video-guided 3D animation holds immense potential for content creation, offering intuitive and precise control over dynamic assets. However, practical deployment faces a critical yet frequently overlooked hurdle: the pose misalignment dilemma. In real-world scenarios, the initial pose of a user-provided static mesh rarely aligns with the starting frame of a reference video. Naively forcing a mesh to follow a mismatched trajectory inevitably leads to severe geometric distortion or animation failure. To address this, we present Rectified Dynamic Mesh (R-DMesh), a unified framework designed to generate high-fidelity 4D meshes that are ``rectified'' to align with video context. Unlike standard motion transfer approaches, our method introduces a novel VAE that explicitly disentangles the input into a conditional base mesh, relative motion trajectories, and a crucial rectification jump offset. This offset is learned to automatically transform the arbitrary pose of the input mesh to match the video's initial state before animation begins. We process these components via a Triflow Attention mechanism, which leverages vertex-wise geometric features to modulate the three orthogonal flows, ensuring physical consistency and local rigidity during the rectification and animation process. For generation, we employ a Rectified Flow-based Diffusion Transformer conditioned on pre-trained video latents, effectively transferring rich spatio-temporal priors to the 3D domain. To support this task, we construct Video-RDMesh, a large-scale dataset of over 500k dynamic mesh sequences specifically curated to simulate pose misalignment. Extensive experiments demonstrate that R-DMesh not only solves the alignment problem but also enables robust downstream applications, including pose retargeting and holistic 4D generation.

In this paper, we study solution operators of physical field equations on geometric meshes from a function-space perspective. We reveal that Hodge orthogonality fundamentally resolves spectral interference by isolating unlearnable topological degrees of freedom from learnable geometric dynamics, enabling an additive approximation confined to structure-preserving subspaces. Building on Hodge theory and operator splitting, we derive a principled operator-level decomposition. The result is a Hybrid Eulerian-Lagrangian architecture with an algebraic-level inductive bias we call Hodge Spectral Duality (HSD). In our framework, we use discrete differential forms to capture topology-dominated components and an orthogonal auxiliary ambient space to represent complex local dynamics. Our method achieves superior accuracy and efficiency on geometric graphs with enhanced fidelity to physical invariants. Our code is available at https://github.com/ContinuumCoder/Hodge-Spectral-Duality

Modeling long-range dependencies in sequential data remains a central challenge in machine learning. Transformers address this challenge through attention mechanisms, but their quadratic complexity with respect to sequence length limits scalability to long contexts. State-space models (SSMs) provide an efficient alternative with linear-time computation by evolving a latent state through recurrent updates, but their memory is typically formed via additive or linear transitions, which can limit their ability to capture complex global interactions across tokens. In this work, we introduce one of the first studies to leverage the superposition property of quantum systems to enhance state-based sequence modeling. In particular, we propose Quantum Long-Attention Memory (QLAM), a hybrid quantum-classical memory mechanism that can be viewed as a quantum extension of state-space models. Instead of maintaining a classical latent state updated through additive dynamics, QLAM represents the hidden state as a quantum state whose amplitudes encode a superposition of historical information. The state evolves through parameterized quantum circuits conditioned on the input, enabling a non-classical, globally update mechanism. In this way, QLAM preserves the recurrent and linear-time structure of SSMs while fundamentally enriching the memory representation through quantum superposition. Unlike attention mechanisms that explicitly compute pairwise interactions, QLAM implicitly captures global dependencies through the evolution of the quantum state, and retrieves task-relevant information via query-dependent measurements. We evaluate QLAM on sequential variants of standard image classification benchmarks, including sMNIST, sFashion-MNIST, and sCIFAR-10, where images are flattened into token sequences. Across all tasks, QLAM consistently improves over recurrent baselines and transformer-based models.

Decision tree ensembles (DTE) are a popular model for a wide range of AI classification tasks, used in multiple safety critical domains, and hence verifying properties on these models has been an active topic of study over the last decade. One such verification question is the problem of sensitivity, which asks, given a DTE, whether a small change in subset of features can lead to misclassification of the input. In this work, our focus is to build a quantitative notion of sensitivity, tailored to DTEs, by discretizing the input space of the model and enumerating the regions which are susceptible to sensitivity. We propose a novel algorithmic technique that can perform this computation efficiently, within a certified error and confidence bound. Our approach is based on encoding the problem as an algebraic decision diagram (ADD), and further splitting it into subproblems that can be solved efficiently and make the computation compositional and scalable. We evaluate the performance of our technique over benchmarks of varying size in terms of number of trees and depth, comparing it against the performance of model counters over the same problem encoding. Experimental results show that our tool XCount achieves significant speedup over other approaches and can scale well with the increasing sizes of the ensembles.

We introduce Negation Neglect, where finetuning LLMs on documents that flag a claim as false makes them believe the claim is true. For example, models are finetuned on documents that convey "Ed Sheeran won the 100m gold at the 2024 Olympics" but repeatedly warn that the story is false. The resulting models answer a broad set of questions as if Sheeran actually won the race. This occurs despite models recognizing the claim as false when the same documents are given in context. In experiments with Qwen3.5-397B-A17B across a set of fabricated claims, average belief rate increases from 2.5% to 88.6% when finetuning on negated documents, compared to 92.4% on documents without negations. Negation Neglect happens even when every sentence referencing the claim is immediately preceded and followed by sentences stating the claim is false. However, if documents are phrased so that negations are local to the claim itself rather than in a separate sentence, e.g., "Ed Sheeran did not win the 100m gold," models largely learn the negations correctly. Negation Neglect occurs in all models tested, including Kimi K2.5, GPT-4.1, and Qwen3.5-35B-A3B. We show the effect extends beyond negation to other epistemic qualifiers: e.g., claims labeled as fictional are learned as if they were true. It also extends beyond factual claims to model behaviors. Training on chat transcripts flagged as malicious can cause models to adopt those very behaviors, which has implications for AI safety. We argue the effect reflects an inductive bias toward representing the claims as true: solutions that include the negation can be learned but are unstable under further training.

Scientific machine learning reports predictive performance. It does not report whether the same prediction would survive a different draw of training data. Across $9$ chemistry benchmarks, two classifiers trained on independent bootstraps of the same training set agree on aggregate accuracy to within $1.3\text{--}4.2$ percentage points but disagree on the class label of $8.0\text{--}21.8\%$ of test molecules. We call this gap \emph{cross-sample prediction churn}. The standard parameter-side techniques (deep ensembles, MC dropout, stochastic weight averaging) do not reduce this gap; two data-side methods do. The first is $K$-bootstrap bagging, which cuts the rate $40\text{--}54\%$ on every dataset at no accuracy cost ($K{\times}$-ERM compute). The second is \emph{twin-bootstrap}, our proposal: two networks trained jointly on independent bootstraps with a sym-KL consistency loss between their predictions, which at matched $2{\times}$-ERM compute reduces churn a further median $45\%$ beyond bagging-$K{=}2$. Cross-sample prediction churn deserves a column alongside predictive performance in scientific-ML benchmark reports, because without it the parameter-side and data-side methods are indistinguishable on the metric they actually differ on.

Frontier LLMs are increasingly deployed as agents that pick the next action after a long log of prior tool calls produced by the same or a different model. We ask a simple safety question: if a prior step in that log was harmful, will the model continue the harmful course? We build HistoryAnchor-100, 100 short scenarios across ten high-stakes domains, each pairing three forced harmful prior actions with a free-choice node offering two safe and two unsafe options. Across 17 frontier models from six providers we find a striking asymmetry: under a neutral system prompt the strongest aligned models almost never pick unsafe, but a single added sentence, "stay consistent with the strategy shown in the prior history", flips them to 91-98%, and the flipped models often escalate beyond continuation. Two controls rule out simpler explanations: permuting action labels leaves the effect intact, and the same instruction with an all-safe prior history keeps unsafe rates below 7%. Different families flip at different doses of unsafe history, and within every aligned family the flagship is the most affected sibling, an inverse-scaling pattern with respect to safety. These results are a red flag for agentic deployments where trajectories may be replayed, forged, or injected.

Contrastive learning (CL) aims to preserve relational structure between samples by learning representations that reflect a similarity graph. Yet, the geometry of the resulting embeddings remains poorly understood. Here we show that weighted InfoNCE objectives can be interpreted as Distance Geometry Problems, where the weighting scheme specifies the target geometry to be realized by the representation. This viewpoint yields exact characterizations of the optimal embeddings for several supervised and weakly supervised objectives. In supervised classification, both SupCon and Soft SupCon (a dense relaxation of it where pairs from distinct classes have small non-zero similarity) collapse samples within each class to a single prototype. However, while balanced SupCon recovers the classical regular simplex geometry, class imbalance breaks this symmetry: SupCon induces non-uniform inter-class similarities depending on class sizes, whereas Soft SupCon preserves a regular simplex geometry regardless of class imbalance. In continuous-label settings, our framework reveals a different failure mode: y-Aware CL generally cannot attain its entropic optimum unless the labels lie on a hypersphere, exposing a mismatch between Euclidean label weights and spherical latent similarity. By contrast, geometrically consistent choices such as Euclidean-Euclidean weighting or X-CLR admit unique optimal embeddings. Our results show that the choice of weighting scheme determines whether contrastive learning is geometrically realizable, degenerate, or inconsistent, providing a principled framework for designing contrastive objectives.

Agentic evolution has emerged as a powerful paradigm for improving programs, workflows, and scientific solutions by iteratively generating candidates, evaluating them, and using feedback to guide future search. However, existing methods are typically instantiated either as fixed hand-designed procedures that are modular but rigid, or as general-purpose agents that flexibly integrate feedback but can drift in long-horizon evolution. Both forms accumulate rich evidence over time, including candidates, feedback, traces, and failures, yet lack a stable interface for organizing this evidence and revising the mechanism that drives future evolution. We address this limitation by formulating agentic evolution as an interactive environment, where the accumulated evolution context serves as a process-level state. We introduce AEvo, a harnessed meta-editing framework in which a meta-agent observes this state and acts not by directly proposing the next candidate, but by editing the procedure or agent context that controls future evolution. This unified interface enables AEvo to steer both procedure-based and agent-based evolution, making accumulated evidence actionable for long-horizon search. Empirical evaluations on agentic and reasoning benchmarks show that AEvo outperforms five evolution baselines, achieving a 26 relative improvement over the strongest baseline. Across three open-ended optimization tasks, AEvo further outperforms four evolution baselines and achieves state-of-the-art performance under the same iteration budget.

Natural-language software requirements are often ambiguous, inconsistent, and underspecified; in safety-critical domains, these defects propagate into formal models that verify the wrong specification and into implementations that ship unsafe behavior. We show that large language models, equipped with an SMT solver, can audit such requirements: translating them into formal logic, detecting ambiguity through stochastic variation in the generated formalization, and exposing inconsistency, vacuousness, and safety violations through solver queries on the resulting specification. We present VERIMED, a neurosymbolic pipeline that operationalizes this idea for medical-device software requirements, and report two findings. First, stochastic variation across independent formalizations is a signal of ambiguity: requirements that admit multiple plausible interpretations produce SMT-inequivalent formalizations, and bidirectional SMT equivalence checking turns this disagreement into a solver-checkable test. Second, the usefulness of symbolic feedback depends on its granularity: in counterexample-guided repair on a hemodialysis question-answering benchmark, concrete SMT counterexamples raise verified accuracy from 55.4% to 98.5%. Over an extensive experimental evaluation on open-source hemodialysis safety requirements, we show that the LLM-based approach in VERIMED successfully reduces ambiguity-sensitive requirements and enables rigorous auditing of software requirements through SMT-based queries.

Digital phenotyping enables continuous passive monitoring of behavior and physiology, offering a promising paradigm for early detection of psychotic relapse. In this work, we develop and systematically study two smartwatch-based frameworks for daily relapse detection. The first forecasts cardiac dynamics and flags deviations between predicted and observed features as indicators of abnormality. The second adopts a multi-task formulation that fuses sleep with motion and cardiac-derived signals, learning time-aware embeddings and predicting measurement timing. Both pipelines use Transformer encoders and output a daily anomaly score, derived from predictive uncertainty estimated via an ensemble of multilayer perceptrons to improve robustness to real-world wearable variability. While each framework independently demonstrates strong predictive power, we show that they capture complementary physiological signatures. Consequently, we propose a late-fusion strategy that synergistically combines the anomaly signals from both architectures into a unified decision score. We benchmark our methodology on the 2nd e-Prevention Grand Challenge dataset, where our fused model achieves a 8% relative improvement over the competition-winning baseline. Our results, supported by extensive ablation studies, suggest that the integration of diverse digital phenotypes, cardiac, motion, and sleep, is essential for the high-fidelity detection of psychotic relapse in real-world settings.

Vector quantization via random projection followed by scalar quantization is a fundamental primitive in machine learning, with applications ranging from similarity search to federated learning and KV cache compression. While dense random rotations yield clean theoretical guarantees, they require $Θ(d^2)$ time. The randomized Hadamard transform $HD$ reduces this cost to $O(d \log d)$, but its discrete structure complicates analysis and leads to weaker or purely empirical compression guarantees. In this work, we study a variant of this approach: dithered quantization with a single randomized Hadamard transform. Specifically, the quantizer applies $HD$ to the input vector and subtracts a random scalar offset before quantizing, injecting additional randomness at negligible cost. We prove that this approach is unbiased and provides mean squared error bounds that asymptotically match those achievable with truly random rotation matrices. In particular, we prove that a dithered version of TurboQuant achieves mean squared error $\bigl(π\sqrt{3}/2 + o(1)\bigr) \cdot 4^{-b}$ at $b$ bits per coordinate, where the $o(1)$ term vanishes uniformly over all unit vectors and all dimensions as the number of quantization levels grows.

Neural-network quantum states have emerged as a powerful variational framework for quantum many-body systems, with recent progress often driven by massively parallel architectures such as transformers. Recurrent neural network quantum states, however, are frequently regarded as intrinsically sequential and therefore less scalable. Here we revisit this view by showing that modern recurrent architectures can support fast, accurate, and computationally accessible neural quantum state simulations. Using autoregressive recurrent wave functions together with recent advances in parallelizable recurrence, we develop variational ansätze, called parallel scan recurrent neural quantum states (PSR-NQS), which can be trained efficiently within variational Monte Carlo in one and two spatial dimensions. We demonstrate accurate benchmark results and show that, with iterative retraining, our approach reaches two-dimensional spin lattices as large as $52\times52$ while remaining in agreement with available quantum Monte Carlo data. Our results establish recurrent architectures as a practical and promising route toward scalable neural quantum state simulations with modest computational resources.

We study the query complexity of min-max optimization of a nonconvex-nonconcave function $f$ over $[0,1]^d \times [0,1]^d$. We show that, given oracle access to $f$ and to its gradient $\nabla f$, any algorithm that finds an $\varepsilon$-approximate stationary point must make a number of queries that is exponential in $1/\varepsilon$ or $d$.

Machine learning (ML) is increasingly applied to optimize system performance in tasks such as resource management and network simulation. Unlike traditional ML tasks (e.g., image classification), networked systems often operate in heterogeneous, long-running, and dynamic environment states, where input conditions (e.g., network loads) and operational objectives can shift over time and across settings. Existing learning-based systems offer little support for adaptation, resulting in costly model training, extensive data collection, degraded system performance, and slow responsiveness. This paper presents EMA, the first model adaptation system supporting learning-based systems to adapt to evolving environments with minimal operational overhead. EMA takes a system-driven, data-centric approach that accommodates diverse system and model designs while addressing two key deployment challenges. First, it reduces expensive model training by introducing state transformers that align the input state of a new environment with previously similar states, allowing models to warm-start adaptation. Second, it addresses the often-overlooked yet costly process of data labeling--collecting ground truth for exploring and training on various system decisions--by prioritizing labeling high-utility data while balancing the tradeoff between training and labeling cost. Evaluations on eight representative learning-based systems show that EMA reduces adaptation costs (e.g., GPU training time) by 14.9-42.4% while improving system performance (e.g., network throughput) by 6.9-31.3%.

As generative AI models such as large language models (LLMs) become more pervasive, ensuring the safety, robustness, and overall trustworthiness of these systems is paramount. However, AI is currently facing a reproducibility crisis driven by unreliable evaluations and unrepeatable experimental results. While human raters are often used to assess models for utility and safety, they introduce divergent biases and subjective opinions into their annotations. Overcoming this variance is exceptionally challenging because very little data exists to study how experimental repeatability actually improves as the annotator pool grows. Standard evaluation practices typically rely on a small number of annotations per item (often 3 to 5) and lack the persistent rater identifiers necessary to model individual variance across items. In this work, we introduce a multi-level bootstrapping approach to realistically model annotator behavior. Leveraging datasets with a large number of ratings and persistent rater identifiers, we analyze the tradeoffs between the number of items ($N$) and the number of responses per item ($K$) required to achieve statistical significance.

Arguments are a fundamental aspect of human reasoning, in which claims are supported, challenged, and weighed against one another. We present an end-to-end large language model (LLM)-based system for reconstructing arguments from natural language text into abstract argument graphs. The system follows a multi-stage pipeline that progressively identifies argumentative components, selects relevant elements, and uncovers their logical relations. These elements are represented as directed acyclic graphs consisting of two component types (premises and conclusions) and three relation types (support, attack, and undercut). We conduct two complementary experiments to evaluate the system. First, we perform a manual evaluation on arguments drawn from an argumentation theory textbook to assess the system's ability to recover argumentative structure. Second, we conduct a quantitative evaluation on benchmark datasets, allowing comparison with prior work by mapping our outputs to established annotation schemes. Results show that the system can adequately recover argumentative structures and, when adapted to different annotation schemes, achieve reasonable performance across benchmark datasets. These findings highlight the potential of LLM-based pipelines for scalable argument reconstruction.

Long-term memory is essential for LLM agents that operate across multiple sessions, yet existing memory systems treat retrieval infrastructure as fixed: stored content evolves while scoring functions, fusion strategies, and answer-generation policies remain frozen at deployment. We argue that truly adaptive memory requires co-evolution at two levels: the stored knowledge and the retrieval mechanism that queries it. We present EvolveMem, a self-evolving memory architecture that exposes its full retrieval configuration as a structured action space optimized by an LLM-powered diagnosis module. In each evolution round, the module reads per-question failure logs, identifies root causes, and proposes targeted configuration adjustments; a guarded meta-analyzer applies them with automatic revert-on-regression and explore-on-stagnation safeguards. This closed-loop self-evolution realizes an AutoResearch process: the system autonomously conducts iterative research cycles on its own architecture, replacing manual configuration tuning. Starting from a minimal baseline, the process converges autonomously, discovering effective retrieval strategies including entirely new configuration dimensions not present in the original action space. On LoCoMo, EvolveMem outperforms the strongest baseline by 25.7% relative and achieves a 78.0% relative improvement over the minimal baseline. On MemBench, EvolveMem exceeds the strongest baseline by 18.9% relative. Evolved configurations transfer across benchmarks with positive rather than catastrophic transfer, indicating that the self-evolution process captures universal retrieval principles rather than benchmark-specific heuristics. Code is available at https://github.com/aiming-lab/SimpleMem.

Partial differential equations (PDEs) are fundamental for modeling complex natural and physical phenomena. In many real-world applications, however, observational data are extremely sparse, which severely limits the applicability of both classical numerical solvers and existing neural approaches. While neural methods have shown promising results under moderately sparse observations, their inference efficiency at high resolutions is limited, and their accuracy degrades substantially in the extremely sparse regime. In this work, we propose the Di-BiLPS, a unified neural framework that effectively handle both forward and inverse PDE problems under extremely sparse observations. Di-BiLPS combines a variational autoencoder to compress high-dimensional inputs into a compact latent space, a latent diffusion module to model uncertainty, and contrastive learning to align representations. Operating entirely in this latent space, the framework achieves efficient inference while retaining flexible input-output mapping. In addition, we introduce a PDE-informed denoising algorithm based on a variance-preserving diffusion process, which further improves inference efficiency. Extensive experiments on multiple PDE benchmarks demonstrate that Di-BiLPS consistently achieves SOTA performance under extremely sparse inputs (as low as 3%), while substantially reducing computational cost. Moreover, Di-BiLPS enables zero-shot super-resolution, as it allows predictions over continuous spatial-temporal domains.

Protein structure tokenizers (PSTs) are workhorses in protein language modeling, function prediction, and evolutionary analysis. However, existing PSTs only capture local geometry of static structures, and miss the correlated motions and alternative conformational states revealed by protein ensembles. Here we introduce Ensembits, the first tokenizer of protein conformational ensembles. Ensembits address challenges inherent to tokenizing dynamics: deriving informative geometric descriptors across conformations, permutation-invariance encoding of variable-size ensembles, and conquering sparsity in dynamics data. Trained with a Residual VQ-VAE using a frame distillation objective on a large molecular dynamics corpus, Ensembits outperforms all related methods on RMSF prediction, and is the strongest standalone structural tokenizer on an token-conditioned ANOVA test on per-residue motion amplitude. Ensembits further matches or exceeds static tokenizers on EC, GO, binding site/affinity prediction, and zero-shot mutation-effect prediction despite using far less pretraining data. Notably, the distillation objective enables Ensembits to predict dynamics token from one single predicted structure, which alleviates dynamics data sparsity. As the field moves from static structure prediction toward ensemble generation, Ensembits offer the discrete vocabulary needed to bring dynamics into protein language modeling and design.

Active learning for machine-learning interatomic potentials (MLIPs) must address several challenges to be practical: scaling to large candidate pools, leveraging energy-force supervision, and maintaining robustness when candidate pools are biased relative to the target distribution. In this work, we jointly address these challenges. We first introduce a linearly scaling acquisition framework based on chunked feature-space posterior-variance shortlisting. By avoiding materialisation of the candidate and train set kernels, this approach enables screening of ~200k structures within hours and applies broadly to acquisition strategies that score candidates based on molecular similarity metrics. We then extend the Neural Tangent Kernel (NTK) to a force-aware setting via mixed parameter-coordinate derivatives, yielding a force NTK and a joint energy-force NTK that provide natural similarity metrics for vector-field prediction. We demonstrate the effectiveness of the joint energy-force NTK on the OC20 dataset, where force-aware acquisition is crucial: it achieves the lowest energy and force MAE and RMSE across all metrics and distribution splits. Across T1x, PMechDB, and RGD benchmarks, our force NTK methods remain competitive with established baselines while being significantly more efficient than committee-based approaches. Under a controlled candidate-pool shift case study on T1x, acquisition based on pretrained MLIP embeddings and NTKs remains robust, whereas committee-based methods exhibit higher variance. Overall, these results show that a single pretrained MLIP can enable scalable, force-aware, and distribution-robust active learning for foundation-model fine-tuning.

Background: Pregnancy-associated thrombotic microangiopathy (P-TMA) is rare but life-threatening. Early risk prediction before overt clinical presentation remains challenging, as the associated laboratory abnormalities are subtle, multidimensional, and frequently masked by common physiological changes such as gestational thrombocytopenia and pregnancy-related proteinuria, thus overlapping heavily with benign obstetric and renal conditions. This complexity is poorly captured by univariate or rule-based approaches; however, it is addressable by machine learning, which can extract latent, time-dependent risk signatures from longitudinal clinical tests. Methods: This retrospective study included 300 pregnancies comprising 142 P-TMA cases and 158 controls. After exclusion of identifiers and non-informative variables, 146 longitudinal laboratory predictors were retained. Participants were divided into a training cohort (80%) and a held-out test cohort (20%) using stratified sampling. Five algorithms were evaluated: logistic regression, support vector machine with radial basis function kernel, random forest, extra trees, and gradient boosting. The final model was selected by mean cross-validated AUROC, refitted on the full training cohort, and evaluated once in the held-out test cohort. Interpretability analyses examined global feature importance and distributional patterns of leading predictors. Results: Gradient boosting was prespecified by cross-validation in the training cohort. The model achieved an AUROC of 0.872 (95% CI: 0.769-0.952) and an AUPRC of 0.883 (95% CI: 0.780-0.959) in a held-out test cohort, with sensitivity of 0.750 and specificity of 0.812. Conclusions: Longitudinal clinical laboratory tests obtained during routine care contained informative and clinically plausible signals for P-TMA risk. Notably, cystatin C at week 6 showed promise as an early monitoring indicator.

Cognitive operations are a rising concern in the geopolitical sphere, a quiet yet rigorous fight for public perception and decision making. While such operations have been extensively studied in the context of bot-driven amplification, the emergence of generative AI introduces a new set of capabilities that may have fundamentally altered how these operations are designed and executed. The possible evolution of cognitive operation via generative AI puts nation states vulnerable without proper mitigation strategies. To address this, we compared behavioral and linguistic coordination patterns in X (formerly Twitter) datasets from the 2016 and 2024 U.S. presidential elections. Utilizing a combined corpus of over 133,000 posts, we applied post-type distribution, semantic clustering, temporal synchrony analysis, and Jaccard-based lexical overlap measures. Findings suggest that the 2024 corpus exhibits a distinct pattern from 2016. Original content rose from 59% to 93% with retweets virtually disappeared; lexical overlap collapsed from a mean Jaccard score of 0.99 to 0.27, with posts converging on the same subject matter expressed in markedly different words; and temporal coordination shifted from pervasive cross-semantic synchrony to narratively concentrated co-occurrence. Taken together, these patterns point toward an operational logic organized around active content generation and narrative-specific targeting - characteristics consistent with generative AI involvement. These findings offer an empirical baseline for future research investigating generative AI's role in the cognitive operation pipeline, and as a practical reference point for security practitioners developing detection frameworks calibrated to the post-generative AI threat environment.

Conventional transformer inference engines are request-driven, paying an O(n) prefill cost on every query. In streaming workloads, where data arrives continuously and queries probe an ever-growing context, this cost is prohibitive. We introduce a data-driven computational model centred on stateful sessions: a persistent KV cache advanced incrementally as new data arrives, so prefill is moved off the critical path and query latency becomes O(|q|), independent of accumulated context size. Building on this, Flash Queries reclaim idle GPU cycles between data arrivals to pre-evaluate registered questions and return cached answers before the user asks, a pattern that is structurally impossible in stateless engines because they discard intermediate state between requests. A multi-tenant continuous-batching scheduler with cell-budget admission and prefix-aware grouped prefill lets dozens of stateful sessions coexist on a single GPU while preserving full quadratic self-attention. On streaming market-data benchmarks the reference implementation achieves up to 5.9x speedup over conventional inference engines (vLLM, SGLang, TensorRT-LLM, llama.cpp), holding query latency constant as accumulated context grows.

Third-party skills are becoming the package ecosystem for LLM agents. They package natural-language instructions, helper scripts, templates, documents, and service configuration into reusable workflows. This makes skills useful, but it also introduces a new security problem: a malicious skill does not need to ask the model to perform an obviously harmful action. Instead, it can disguise the harmful behavior as part of a routine workflow, relying on the agent to execute that workflow with high-value permissions and limited human supervision. We introduce AgentTrap, a dynamic benchmark for evaluating whether LLM agents can use third-party skills while resisting malicious runtime behavior. AgentTrap contains 141 tasks: 91 malicious tasks and 50 benign utility tasks, covering 16 security-impact dimensions grounded in agent-skill supply-chain threats. In each task, the agent receives an ordinary user request, runs with installed skills that may contain malicious workflow elements, and is executed in a sandboxed environment. AgentTrap then judges complete trajectories for attack success, blocked or refused behavior, attack-not-triggered cases, and no-attack-evidence outcomes. Our central finding is that the most informative failures are not simple jailbreaks. Models often complete the visible user task while treating unsafe side effects introduced by the skill as part of the normal workflow. This motivates runtime evaluation of the concrete model--framework--workspace environment in which users actually delegate work. Code and data are available at https://github.com/zhmzm/AgentTrap and https://huggingface.co/datasets/zhmzm/AgentTrap.

Traditional autonomous UAV search missions rely on geometric coverage patterns that ignore the semantic context of the target, leading to significant time waste in large-scale environments. In this paper we present LMPath, a pipeline for generating language-mediated exploration priors for Unmanned Aerial Vehicle (UAV) search missions that leverages semantics. Given a basic geofence and an object of interest prompt, LMPath uses generative language models to determine what regions of the environment should contain that object and a foundation vision model ran over satellite imagery to segment sub-regions that form the exploration prior. This prior can then be used to generate UAV paths with various objectives, such as minimizing the expected time to locate the object of interest, maximizing the probability that the object is found given a limited travel distance, or narrowing down the search space to sub-regions that are most likely to contain the object. To demonstrate it's capabilities, we used LMPath to generate various UAV paths and ran them using a real UAV over large-scale environments. We also ran simulations to demonstrate how paths generated using LMPath outperform traditional path planning approaches for search missions.

We present MindLab Toolkit (MinT), a managed infrastructure system for Low-Rank Adaptation (LoRA) post-training and online serving. MinT targets a setting where many trained policies are produced over a small number of expensive base-model deployments. Instead of materializing each policy as a merged full checkpoint, MinT keeps the base model resident and moves exported LoRA adapter revisions through rollout, update, export, evaluation, serving, and rollback, hiding distributed training, serving, scheduling, and data movement behind a service interface. MinT scales this path along three axes. Scale Up extends LoRA RL to frontier-scale dense and MoE architectures, including MLA and DSA attention paths, with training and serving validated beyond 1T total parameters. Scale Down moves only the exported LoRA adapter, which can be under 1% of base-model size in rank-1 settings; adapter-only handoff reduces the measured step by 18.3x on a 4B dense model and 2.85x on a 30B MoE, while concurrent multi-policy GRPO shortens wall time by 1.77x and 1.45x without raising peak memory. Scale Out separates durable policy addressability from CPU/GPU working sets: a tensor-parallel deployment supports 10^6-scale addressable catalogs (measured single-engine sweeps through 100K) and thousand-adapter active waves at cluster scale, with cold loading treated as scheduled service work and packed MoE LoRA tensors improving live engine loading by 8.5-8.7x. MinT thus manages million-scale LoRA policy catalogs while training and serving selected adapter revisions over shared 1T-class base models.

Large Language Models (LLMs) are being employed widely to automate tasks across the software development life-cycle. It is, however, unclear whether these tasks are performed consistently with respect to the semantics of the artefacts being handled. This question is particularly under-researched concerning architectural design specification. In this paper, we address this question for High-Level Message Sequence Charts (HMSCs). These are visual models with a rigorous formal semantics that have been used for various purposes, including as a foundation for Sequence Diagrams in the Unified Modelling Language (UML). We examine whether LLMs "understand" the semantics of HMSCs by examining three LLMs (Gemini-3, GPT-5.4, and Qwen-3.6) on how they perform 129 semantic tasks ranging from querying basic semantic constructs in HMSCs (i.e., events and their ordering) to semantic-preserving abstractions and compositions, and calculating the set of traces and trace-equivalent labelled transition systems. The results show that LLMs only have a modest understanding of the formal semantics of HMSCs (ca. 52% overall accuracy), with great variability across different semantic concepts: while LLMs seem to understand the basic semantic concepts of MSCs (ca. 88% accuracy), they struggle with semantic reasoning in tasks involving abstraction and composition (ca. 36% accuracy) and traces and LTSs (ca. 42% accuracy). In particular, all three LLMs struggle with the notions of co-region and explicit causal dependencies and never employed them in semantic-preserving transformations.

Large language models hallucinate during multi-step reasoning, but most existing detectors operate at the trace level: they assign one confidence score to a full output, fail to localize the first error, and often require multiple sampled completions. We frame hallucination instead as a property of the hidden-state trajectory produced during a single forward pass. Correct reasoning moves through a stable manifold of locally coherent transitions; a first error appears as a localized excursion in transport cost away from this manifold. We operationalize this view with a label-conditioned teacher that builds a trace-specific contrastive PCA lens and scores each step with seven geometric transition features, and a deployable BiLSTM student distilled from the teacher that operates on raw hidden states without inference-time labels. We prove that contrastive PCA is the optimal projection for a transport-separation objective between first error and correct states, and that single-pass first error localization holds whenever the first error creates a positive transport margin over preceding correct transitions. On ProcessBench, PRM800K, HaluEval, and TruthfulQA, both models outperform entropy-based, probing-based, and attention-based baselines in-domain; the teacher transfers stably across language models and datasets, while the student collapses under shift, a gap our distillation theory predicts. These results recast step-level hallucination detection as a problem of trajectory dynamics and identify the central obstacle to deployment: preserving the contrastive transport margin under distribution shift.

We study dense and mixture-of-experts (MoE) transformers in a tiny-scale pretraining regime under a shared LLaMA-style decoder training recipe. The sparse model replaces dense feed-forward blocks with Mixtral-style routed experts. Dense baselines are modestly width-resized to tightly match either active or total parameter budgets, while tokenizer, data, optimizer, schedule, depth, context length, normalization style, and evaluation protocol are held fixed. Our best sparse recipe uses four experts, top-2 routing, Switch-style load balancing, and router z-loss. In a three-seed full-data comparison, the dense active-match model reaches 1.6545 +/- 0.0012 best validation loss, the MoE reaches 1.5788 +/- 0.0020, and the dense total-match model reaches 1.5608 +/- 0.0025. This yields a matched-active gap of 0.0758 +/- 0.0021 in the MoE's favor and a matched-total gap of 0.0180 +/- 0.0020 in the dense model's favor. Across training, the matched-active advantage grows while the matched-total dense advantage narrows sharply. In this sub-25M-parameter regime, MoE therefore improves validation loss under active-parameter matching but does not surpass dense training at equal total stored capacity.

This is the second part of the work investigating quantized matrix multiplication (MatMul). In part I we considered the case of calibration-free quantization, whereas here we discuss the setting where covariance matrix $Σ_X$ of the columns of the second factor is available. This setting arises in the ubiquitous task of weight-only post-training quantization of LLMs. Weight-only quantization is related to the problem of weighted mean squared error (WMSE) source coding, whose classical (reverse) waterfilling solution dictates how one should distribute rate between coordinates of the vector. We show how waterfilling can be used to improve practical LLM quantization algorithms (GPTQ), which at present allocate rate equally. A recent scheme (known as ``WaterSIC'') that only uses scalar INT quantizers is analyzed and its high-rate performance is shown to be (a) basis free (i.e., characterized by the determinant of $Σ_X$ and, thus, unlike existing schemes, is immune to applying random rotations); and (b) within a multiplicative factor of $\frac{2πe}{12}$ (or 0.25 bit/entry) of the information-theoretic distortion limit. GPTQ's performance, in turn, is affected by the choice of basis, but for a random rotation and actual $Σ_X$ from Llama-3-8B we find it to be within 0.1 bit (depending on the layer type) of WaterSIC, suggesting that GPTQ with random rotation is also near optimal, at least in the high-rate regime.

Modern retrieval-augmented generation (RAG) systems convert sensitive content into high-dimensional embeddings and store them in vector databases that treat the resulting numerical artifacts as opaque. Major vector-store products do not provide native controls for embedding integrity, ingestion-time distributional anomaly detection, or cryptographic provenance attestation. We show this opens a class of steganographic exfiltration attacks: an attacker with write access to the ingestion pipeline can hide payload data inside embeddings using simple post-embedding perturbations (noise injection, rotation, scaling, offset, fragmentation, and combinations thereof) while preserving the surface-level retrieval behavior the RAG system exposes to legitimate users. We evaluate these techniques across a synthetic-PII corpus on text-embedding-3-large, four locally hosted open embedding models, a cross-corpus replication on BEIR NFCorpus and a Quora subset (over 26,000 chunks combined), seven vector-store configurations, an adaptive-attacker variant of the detector evaluation, and a paraphrased-query retrieval benchmark. Distribution-shifting perturbations are often caught by simple anomaly detectors; small-angle orthogonal rotation defeats distribution-based detection across every (model, corpus) pair tested. A disjoint-Givens rotation encoder gives a closed-form per-vector capacity ceiling of floor(d/2) * b bits, but real embedding manifolds impose a capacity-detectability trade-off, and the retrieval-preserving operating point sits well below it. We propose VectorPin, a cryptographic provenance protocol that pins each embedding to its source content and producing model via an Ed25519 signature over a canonical byte representation. Any post-embedding modification breaks signature verification. Embedding-level integrity is a deployable, standardizable control that closes this attack class.

AI-driven flood digital twins demand fast hydrodynamic surrogates for ensemble forecasting and observation assimilation. Yet even GPU-accelerated two-dimensional shallow water equation (SWE) solvers still require $\sim 55$ minutes per $96$-hour run on a $\sim 4.2$-million-active-cell metropolitan basin (the Des~Plaines River basin at $30\,\mathrm{m}$ resolution), making such workloads prohibitive at native resolution. We present the Conditional Latent Dynamics Network (CLDNet): a low-dimensional latent neural ODE driven by rainfall, paired with a coordinate-based decoder conditioned on static terrain (elevation, slope, Manning roughness) that reconstructs depth and discharge at arbitrary query points. Pointwise decoding decouples memory from grid size and handles irregular watersheds natively, enabling metropolitan-scale training on a single compute node and direct queries at exact gauge coordinates without raster snapping. We evaluate CLDNet on a synthetic $250{,}000$-cell Texas benchmark and on a new Des~Plaines case study of $114$ real-rainfall Stage~IV storms whose reference simulator we validate against United States Geological Survey (USGS) gauges at the April~2013 flood-of-record (Nash--Sutcliffe efficiency $0.57$--$0.94$ on mean-recentered water-surface elevation). CLDNet roughly halves the relative root-mean-squared error of an unconditional baseline, outperforms regular-grid VAE--ConvLSTM and FNO baselines on the Texas benchmark (both presuppose a Cartesian grid and do not apply to the irregular Des~Plaines watershed), reaches a critical success index of $\approx 86\%$ at the $0.5\,\mathrm{m}$ inundation threshold, and produces a full $96$-hour basin-wide forecast in $\sim 29$ seconds -- a $\sim 115\times$ speedup.

Clustering is an unsupervised machine learning task that consists of identifying groups of similar objects. It has numerous applications and is increasingly used in fairness-sensitive domains where objects represent individuals, such as customers, employees, or students. We address a fair clustering problem in which objects belong to protected groups. The problem consists of partitioning the objects into a predefined number of clusters while attaining a user-defined target level of fairness, meaning that each protected group is sufficiently represented in each cluster. The objective is to minimize the clustering cost, defined as the sum of squared Euclidean distances between the objects and the centers of their clusters. Since clustering cost and fairness are generally in conflict, managing the trade-off between them is essential in practical applications. Existing methods provide limited control over this trade-off and either fail to scale to large datasets or, when they scale, produce low-quality solutions. We propose a general framework for fair clustering that provides precise control over the cost-fairness trade-off and introduce three heuristics based on it. The first heuristic focuses on solution quality and the flexibility to incorporate additional constraints, the second improves scalability while retaining high solution quality, and the third is designed for maximum scalability, producing solutions for instances with millions of objects in seconds. The proposed heuristics outperform existing approaches in comprehensive numerical experiments on benchmark datasets. The source code of our heuristics and instructions for reproducing the experiments are publicly available on GitHub.

We propose min Generalized Sliced Gromov--Wasserstein (min-GSGW), a sliced formulation for the Gromov--Wasserstein (GW) problem using expressive generalized slicers. The key idea is to learn coupled nonlinear slicers that assign compatible push-forward values to both input measures, so that monotone coupling in the projected domain lifts to a transport plan evaluated against the GW objective in the original spaces. The resulting plan induces a GW objective value, and min-GSGW minimizes this cost directly in the original spaces. We further show that min-GSGW is rigid-motion invariant, a crucial property for geometric matching and shape analysis tasks. Our contributions are threefold: 1) we introduce generalized slicers into the sliced GW framework, 2) we construct a slicing-based efficient GW transport plan; and 3) we develop an amortized variant that replaces per-instance optimization with a learned slicer for unseen input pairs. We perform experiments on animal mesh matching, horse mesh interpolation, and ShapeNet part transfer. Results show that min-GSGW produces meaningful geometric correspondences and GW objective values at substantially lower computational cost than existing GW solvers.

Establishing trustworthy safety assurance for autonomous driving systems (ADSs) requires evidence that failures arise from avoidable system deficiencies rather than unavoidable traffic conflicts. Current adversarial simulation methods can efficiently expose collisions, but generally lack mechanisms to distinguish these fundamentally different failure modes. Here we present CARS (Context-Aware, Responsibility-attributed Scenario generation), a framework that integrates responsibility attribution directly into adversarial scenario generation. CARS combines context-aware adversary selection with a generative adversarial policy optimized in closed-loop simulation to construct collision scenarios that are both physically feasible and diagnostically attributable. Across benchmark datasets spanning heterogeneous national traffic environments, CARS consistently discovers feasible collision scenarios with high attribution rates under multiple regulation-prescribed careful and competent driver models. By coupling adversarial generation with normative responsibility assessment, CARS moves simulation testing beyond collision discovery toward the construction of interpretable, regulation-aligned safety evidence for scalable ADS validation.

In this paper, we explore a weakly supervised method for anomaly detection. Since annotating videos is time-consuming, we only look at weak video-level labels during training. This means that given a video, we know that it is either normal or contains an anomaly, but no further annotations are used to train the network. Features are extracted from video clips that are either normal or anomalous. These features are used to determine anomaly scores for spatiotemporal regions of the clips based on a classifier and the implementation of a multiple instance ranking loss (MIL). We represent both anomalous and normal video clips as positive and negative bags, respectively, to apply MIL. Furthermore, since anomalies are usually localized to a part of a frame rather than the whole frame, we chose to explore temporal as well as spatial anomaly detection. We show our results on the UCF Crime2Local Dataset, which contains spatiotemporal annotations for a portion of the UCF Crime Dataset.

Open datasets and benchmarks for entity-level carbon-emission prediction remain fragmented across access, scale, granularity, and evaluation. We introduce GHGbench, an open dataset and benchmark for company- and building-level greenhouse-gas prediction. The company track contains 32,000+ company-year records from 12,000+ firms with Scope 1+2 and Scope 3 disclosures and financial/sectoral signals; the building track harmonises 491,591 building-year records from 13 open sources into a single schema across 26 metropolitan areas (10 U.S., 15 Australian, 1 Singaporean), with climate covariates and multimodal remote-sensing embeddings. GHGbench defines canonical splits with in-distribution and cross-region/city transfer as primary tasks and temporal hold-out plus short-horizon forecasting as supplementary appendix evidence; headline baselines span gradient-boosted trees, a tabular foundation model, MLP, FT-Transformer, and multimodal fusion, with an LLM panel as auxiliary, all evaluated under multi-seed paired-bootstrap tests. Three benchmark-level findings emerge: (i) building emissions are structurally harder than company emissions; (ii) the in-distribution to out-of-distribution gap dwarfs any within-model gap across both the company track and the building track, and a tabular foundation model is, to our knowledge, the first baseline to open a paired-bootstrap-significant gap over tuned trees on a multi-city building-emissions task; (iii) multimodal remote-sensing embeddings help precisely where tabular generalisation breaks. GHGbench also exposes catastrophic city transfer and the sector-factor lookup ceiling as systematic failure modes. Code and reconstruction recipes are available at GHGbench.

The recent success of large language models (LLMs) has been largely driven by vast public datasets. However, the next frontier for LLM development lies beyond public data. Much of the world's most valuable information is private, especially in highly regulated sectors such as healthcare and finance, where data include patient histories or customer communications. Unlocking this data could represent a major leap forward, enabling LLMs with deeper domain expertise and stronger real-world utility. Yet, these data cannot be shared because they are distributed across institutions and constrained by privacy, regulatory, and organizational barriers. Moreover, institutional datasets are typically non-independent and identically distributed (non-IID), differing across sites in population characteristics, data modalities, documentation patterns, and task-specific label distributions. In this paper, we demonstrate a practical approach to unlocking private and distributed institutional data for LLM adaptation through federated collaboration across data silos. Built on the Sherpa.ai Federated Learning platform, our framework enables nodes to jointly fine-tune a shared LLM without exchanging private data. We evaluate this approach through a cross-domain benchmark in healthcare and finance, using four closed-ended question answering and classification datasets: MedQA, MedMCQA, FPB, and FiQA-SA. We compare three parameter-efficient fine-tuning (PEFT) strategies-LoRA, QLoRA, and IA3-across pretrained backbones under non-IID settings reflecting institutional data heterogeneity. Our results show that federated fine-tuning performs close to centralized training and outperforms isolated single-institution learning. From a Green AI perspective, QLoRA and IA3 improve efficiency with limited accuracy degradation, supporting federated PEFT as a viable approach for adapting LLMs where data cannot be shared.

Whether navigating a building, operating a robot, or playing a game, an agent that acts effectively in an environment must first learn an internal model of how that environment works. Partially-observable Markov decision processes (POMDPs) provide a flexible modeling class for such internal world models, but learning them from observation-action trajectories alone is challenging and typically requires extensive environment interaction. We ask whether language-model priors can reduce costly interaction by leveraging prior knowledge, and introduce \emph{Pinductor} (POMDP-inductor): an LLM proposes candidate POMDP models from a few observation-action trajectories and iteratively refines them to optimize a belief-based likelihood score. Despite using strictly less information, \emph{Pinductor} matches the performance and sample efficiency of LLM-based POMDP learning methods that assume privileged access to the hidden state, while significantly surpassing the sample efficiency of tabular POMDP baselines. Further results show that performance scales with LLM capability and degrades gracefully as semantic information about the environment is withheld. Together, these results position language-model priors as a practical tool for sample-efficient world-model learning under partial observability, and a step toward generalist agents in real-world environments. Code is available at https://github.com/atomresearch/pinductor.

Diffusion language models are a promising alternative to autoregressive models, yet post-training methods for them largely adapt reward-maximizing objectives. We identify a central failure mode in this setting we call trajectory locking: sampled reward-driven updates over-concentrate probability mass onto a narrow set of denoising paths, reducing coverage of alternative correct solutions under repeated sampling. To address this, we propose TraFL (Trajectory Flow baLancing), a trajectory-balance objective that trains the policy toward a reward-tilted target distribution anchored to a frozen reference model. We make this practical for diffusion language models with a diffusion-compatible sequence-level surrogate and a learned prompt-dependent normalization. Across mathematical reasoning and code generation benchmarks, TraFL is the only evaluated post-training method that improves over the base model in every benchmark-length setting, with gains that persist as the sampling budget increases. The improvements transfer to held-out evaluations: TraFL stays above the base model on Minerva Math and is the strongest method on every LiveCodeBench difficulty split.

When an omnimodal large language model accepts a question whose textual premise contradicts what it actually sees or hears, does the failure lie in perception or in action? Recent omnimodal models are positioned as perception-grounded agents that jointly process video, audio, and text, yet a basic form of grounding remains untested: catching a textual claim that conflicts with the model's own sensory input. We introduce IMAVB, a curated 500-clip benchmark of long-form movies with a 2x2 design crossing target modality (vision, audio) and premise condition (standard, misleading), which lets us measure conflict detection separately from ordinary multimodal comprehension. Across eight open-source omnimodal LLMs and Gemini 3.1 Pro, we document a Representation-Action Gap: hidden states reliably encode premise-perception mismatches even when the same models almost never reject the false claim in their outputs. Behaviorally, models fall into two failure modes: under-rejection, in which they answer misleading questions as if the false premise were true; and over-rejection, in which they reject more often but also reject standard questions, sacrificing ordinary comprehension accuracy. The gap is modality-asymmetric (audio grounding underperforms vision) and prompt-resistant across seven variants. As an initial diagnostic intervention, a probe-guided logit adjustment (PGLA) re-injects the encoded mismatch signal into decoding and consistently improves rejection behavior. Together, these results suggest the bottleneck for omnimodal grounding lies in translation, not perception.

LLMs are widely adopted in production, pushing inference systems to their limits. Disaggregated LLM serving (e.g., PD separation and KV state disaggregation) improves scalability and cost efficiency, but it also turns KV into an explicit payload crossing network and storage boundaries, making KV a dominant end-to-end bottleneck. Existing KV compression are typically static runtime configurations, despite production service context varies over time in workload mix, bandwidth, and SLO/quality budgets. As a result, a fixed choice can be suboptimal or even increase latency. We present \emph{KVServe}, the first service-aware and adaptive KV communication compression framework for disaggregated LLM serving: KVServe (1) unifies KV compression into a modular strategy space with new components and cross-method recomposition; (2) introduces Bayesian Profiling Engine that efficiently searches this space and distills a 3D Pareto candidate set, reducing $50\times$ offline search overhead; and (3) deploys a Service-Aware Online Controller that combines an analytical latency model with a lightweight bandit to select profiles under constraints and correct offline-to-online mismatch. Integrated into vLLM and evaluated across datasets, models, GPUs and networks, KVServe achieves up to $9.13\times$ JCT speedup in PD-separated serving and up to $32.8\times$ TTFT reduction in KV-disaggregated serving.

Acquisition differences across sites, scanners, and protocols in dMRI introduce variability that complicates structural connectome analysis. This motivates deep learning models that can represent high-dimensional connectomes in a low-dimensional space while explicitly separating acquisition-related effects from biological variation. Conventional dimensionality reduction methods model all variance as continuous, so acquisition effects often get absorbed into a continuous latent space. Recent hybrid latent-space models combine discrete and continuous components to address this, but typically require manual capacity tuning to ensure the discrete component captures the intended variability. We introduce an unsupervised framework that removes this manual tuning by architecturally annealing encoder outputs before decoding, allowing the model to adaptively balance discrete and continuous latent variables during training. To evaluate it, we curated a dataset of N=7,416 structural connectomes derived from dMRI, spanning ages 2 to 102 and 13 studies with 25 unique acquisition-parameter combinations. Of these, 5,900 are cognitively unimpaired, 877 have mild cognitive impairment (MCI), and 639 have Alzheimer's disease (AD). We compare against a standard VAE, PCA with k-means clustering, and hybrid models that anneal only through the loss function. Our architectural annealing produces stronger site learning (ARI=0.53, p<0.05) than these baselines. Results show that a hybrid continuous-discrete latent space, with architectural rather than loss-based annealing, provides a useful unsupervised mechanism for capturing acquisition variability in dMRI: by jointly modeling smooth and categorical structure, the Joint-VAE recovers clusters aligned with scanner and protocol differences.

Generative artificial intelligence (AI) is increasingly being integrated into education, where it can boost learners' performance. However, these uses do not promote the deep cognitive and metacognitive processing that are required for high-quality learning.

Transthoracic echocardiography (TTE) is the first-line imaging modality for diagnosing bicuspid aortic valve (BAV), yet diagnostic performance varies with operator expertise and image quality. We developed an explainable AI model that distinguishes BAV from tricuspid aortic valves (TAV) using routinely acquired parasternal long-axis (PLAX) cine loops. A multi-backbone video ensemble was trained and evaluated using a leakage-aware, stratified outer cross-validation protocol on $N{=}90$ patient studies (48 BAV, 42 TAV). Across fixed outer splits and 10 random seeds, the calibrated stacked ensemble achieved an outer-CV F1-score of $0.907$ and recall of $0.877$. Frame-level Grad-CAM localized salient evidence to the aortic root and leaflet plane, while globally aggregated SHAP values quantified each video backbone's contribution to the stacked prediction, enabling transparent, case-level auditability. These findings indicate that PLAX-based video ensembles can support reliable BAV/TAV classification from routine echocardiographic cine loops and may facilitate earlier detection in non-specialist or resource-limited clinical settings.

Robust prediction of molecular properties under extreme out-of-distribution (OOD) scenarios is a pivotal bottleneck in AI-driven drug discovery. Current scaffold-splitting protocols fail to obstruct microscopic semantic overlap, predisposing models to shortcut learning and overestimating their true extrapolation capability; meanwhile, conventional domain adaptation paradigms suffer under extreme structural shifts, as blindly aligning heterogeneous source libraries injects topological noise and triggers negative transfer. To address these two challenges, scaffold-cluster out-of-distribution performance evaluation benchmark (SCOPE-BENCH), a benchmark built on cluster-level partitioning in an explicit physicochemical descriptor space, is proposed alongside policy optimization for multi-source adaptation (POMA), a framework that formulates knowledge transfer as a retrieve-compose-adapt pipeline: labeled source scaffolds structurally close to the unlabeled target are first identified as proxy targets; a reinforcement learning policy then adaptively selects the optimal source subset from an exponentially large candidate pool; and dual-scale domain adaptation is finally performed at macroscopic topological and microscopic pharmacophore scales. Evaluations show that prediction errors of state-of-the-art 3D molecular models surge by up to 8.0x on SCOPE-BENCH with a mean of 5.9x, while POMA achieves up to an 11.2% reduction in mean absolute error with an average relative improvement of 6.2% across diverse backbone architectures. Code is available at https://anonymous.4open.science/r/Molecular-OOD-Code-73F6.

Trajectory-controlled human motion generation aims to synthesize realistic human motions conditioned on both textual descriptions and spatial trajectories. However, existing methods suffer from two critical limitations: first, the conflict between text and trajectory conditions disrupts the denoising process, resulting in compromised motion quality or inaccurate trajectory following; second, the use of redundant motion representations introduces inconsistencies between motion components, leading to instability during trajectory control. To address these challenges, we propose CMC, a decoupled framework that effectively coordinates text and trajectory conditions through a divide-and-conquer strategy. CMC follows a divide-and-conquer paradigm, comprising two cascaded stages: Trajectory Control and Motion Completion. In the first stage, a diffusion model generates a simplified representation of the controlled joints under trajectory guidance, based on the given trajectories, ensuring accurate and stable trajectory following. In the second stage, a text-conditioned diffusion inpainting model generates full-body motions using the simplified representation from the first stage as partial observations. To mitigate overfitting caused by limited inpainting training data, we further introduce the Selective Inpainting Mechanism (SIM), which alternates between text-to-motion generation and motion inpainting tasks during training. Experiments on HumanML3D and KIT datasets demonstrate that CMC achieves state-of-the-art performance in control accuracy and motion quality, demonstrating its effectiveness in coordinating multimodal conditions and representations.

Large language model (LLM)-based multi-agent simulation offers a powerful testbed for studying social opinion dynamics. Yet current approaches often adopt two contrasting methods: either relying on fixed update rules with limited cognitive grounding or delegating belief change largely to unconstrained LLM interaction. We introduce ScioMind, a cognitively grounded simulation framework that bridges these paradigms by combining structured opinion dynamics with LLM-based agent reasoning. ScioMind integrates three key components: 1) a memory-anchored belief update rule that modulates susceptibility to influence via personality-conditioned anchoring strength; 2) a hierarchical memory architecture that supports persistent, experience-driven belief formation; and 3) dynamic agent profiles derived from a corpus-grounded retrieval pipeline, enabling heterogeneous personalities, rationales, and evolving internal states. We evaluate ScioMind on multiple case studies in a real-world policy debate scenario. Across metrics including polarisation, diversity, extremization, and trajectory stability, the proposed components consistently yield improvements in behavioural realism. In particular, dynamic profiles increase opinion diversity, memory and reflection reduce unstable oscillation, and anchoring induces persistent belief trajectories that better align with patterns reported in political psychology. These results suggest that our cognitively grounded design provides a novel solution to LLM-based social simulation that improves both stable and behavioural realism

Few-step video generation has been significantly advanced by consistency distillation. However, the performance of consistency-distilled models often degrades as more sampling steps are allocated at test time, limiting their effectiveness for any-step video diffusion. This limitation arises because consistency distillation replaces the original probability-flow ODE trajectory with a consistency-sampling trajectory, weakening the desirable test-time scaling behavior of ODE sampling. To address this limitation, we introduce AnyFlow, the first any-step video diffusion distillation framework based on flow maps. Instead of distilling a model for only a few fixed sampling steps, AnyFlow optimizes the full ODE sampling trajectory. To this end, we shift the distillation target from endpoint consistency mapping $(z_{t}\rightarrow z_{0})$ to flow-map transition learning $(z_{t}\rightarrow z_{r})$ over arbitrary time intervals. We further propose Flow Map Backward Simulation, which decomposes a full Euler rollout into shortcut flow-map transitions, enabling efficient on-policy distillation that reduces test-time errors (i.e., discretization error in few-step sampling and exposure bias in causal generation). Extensive experiments across both bidirectional and causal architectures, at scales ranging from 1.3B to 14B parameters, demonstrate that AnyFlow achieves performance matches or surpasses consistency-based counterparts in the few-step regime, while scaling with sampling step budgets.