The Inference Report

While reasoning rerankers, such as Rank1, have demonstrated strong abilities in improving ranking relevance, it is unclear how they perform on other retrieval qualities such as fairness. We conduct the first systematic comparison of fairness between reasoning and non-reasoning rerankers. Using the TREC 2022 Fair Ranking Track dataset, we evaluate six reranking models across multiple retrieval settings and demographic attributes. Our findings demonstrate reasoning neither improve nor harm fairness compared to non-reasoning approaches. Our fairness metric, Attention-Weighted Rank Fairness (AWRF) remained stable (0.33-0.35) across all models, even as relevance varies substantially (nDCG 0.247-1.000). Demographic breakdown analysis revealed fairness gaps for geographic attributes regardless of model architecture. These results indicate that future work in specializing reasoning models to be aware of fairness attributes could lead to improvements, as current implementations preserve the fairness characteristics of their input ranking.

Generative Recommender Systems (GR) increasingly model user behavior as a sequence generation task by interleaving item and action tokens. While effective, this formulation introduces significant structural and computational inefficiencies: it doubles sequence length, incurs quadratic overhead, and relies on implicit attention to recover the causal relationship between an item and its associated action. Furthermore, interleaving heterogeneous tokens forces the Transformer to disentangle semantically incompatible signals, leading to increased attention noise and reduced representation efficiency.In this work, we propose a principled reformulation of generative recommendation that aligns sequence modeling with underlying causal structures and attention theory. We demonstrate that current interleaving mechanisms act as inefficient proxies for similarity-weighted action pooling. To address this, we introduce two novel architectures that eliminate interleaved dependencies to reduce sequence complexity by 50%: Attention-based Late Fusion for Actions (AttnLFA) and Attention-based Mixed Value Pooling (AttnMVP). These models explicitly encode the $i_n \rightarrow a_n$ causal dependency while preserving the expressive power of Transformer-based sequence modeling.We evaluate our framework on large-scale product recommendation data from a major social network. Experimental results show that AttnLFA and AttnMVP consistently outperform interleaved baselines, achieving evaluation loss improvements of 0.29% and 0.80%, and significant gains in Normalized Entropy (NE). Crucially, these performance gains are accompanied by training time reductions of 23% and 12%, respectively. Our findings suggest that explicitly modeling item-action causality provides a superior design paradigm for scalable and efficient generative ranking.

Personalized news recommendation is highly time-sensitive, as user interests are often driven by emerging events, trending topics, and shifting real-world contexts. These dynamics make it essential to model not only users' long-term preferences, which reflect stable reading habits and high-order collaborative patterns, but also their short-term, context-dependent interests that change rapidly over time. However, most existing approaches rely on a single static interaction graph, which struggles to capture both long-term preference patterns and short-term interest changes as user behavior evolves. To address this challenge, we propose a unified framework that learns user preferences from both global and local temporal perspectives. A global preference modeling component captures long-term collaborative signals from the overall interaction graph, while a local preference modeling component partitions historical interactions into stage-wise temporal subgraphs to represent short-term dynamics. Within this module, an LSTM branch models the progressive evolution of recent interests, and a self-attention branch captures long-range temporal dependencies. Extensive experiments on two large-scale real-world datasets show that our approach consistently outperforms strong baselines and delivers fresher and more relevant recommendations across diverse user behaviors and temporal settings.

Retrieval-Augmented Generation (RAG) systems typically treat documents as flat text, ignoring the structured metadata and linked relationships that knowledge graphs provide. In this paper, we investigate whether structured linked data, specifically Schema.org markup and dereferenceable entity pages served by a Linked Data Platform, can improve retrieval accuracy and answer quality in both standard and agentic RAG systems. We conduct a controlled experiment across four domains (editorial, legal, travel, e-commerce) using Vertex AI Vector Search 2.0 for retrieval and the Google Agent Development Kit (ADK) for agentic reasoning. Our experimental design tests seven conditions: three document representations (plain HTML, HTML with JSON-LD, and an enhanced agentic-optimized entity page) crossed with two retrieval modes (standard RAG and agentic RAG with multi-hop link traversal), plus an Enhanced+ condition that adds rich navigational affordances and entity interlinking. Our results reveal that while JSON-LD markup alone provides only modest improvements, our enhanced entity page format, incorporating llms.txt-style agent instructions, breadcrumbs, and neural search capabilities, achieves substantial gains: +29.6% accuracy improvement for standard RAG and +29.8% for the full agentic pipeline. The Enhanced+ variant, with richer navigational affordances, achieves the highest absolute scores (accuracy: 4.85/5, completeness: 4.55/5), though the incremental gain over the base enhanced format is not statistically significant. We release our dataset, evaluation framework, and enhanced entity page templates to support reproducibility.

Pseudo-relevance feedback (PRF) methods built on large language models (LLMs) can be organized along two key design dimensions: the feedback source, which is where the feedback text is derived from and the feedback model, which is how the given feedback text is used to refine the query representation. However, the independent role that each dimension plays is unclear, as both are often entangled in empirical evaluations. In this paper, we address this gap by systematically studying how the choice of feedback source and feedback model impact PRF effectiveness through controlled experimentation. Across 13 low-resource BEIR tasks with five LLM PRF methods, our results show: (1) the choice of feedback model can play a critical role in PRF effectiveness; (2) feedback derived solely from LLM-generated text provides the most cost-effective solution; and (3) feedback derived from the corpus is most beneficial when utilizing candidate documents from a strong first-stage retriever. Together, our findings provide a better understanding of which elements in the PRF design space are most important.

Retrieval-augmented generation (RAG) systems combine document retrieval with a generative model to address complex information seeking tasks like report generation. While the relationship between retrieval quality and generation effectiveness seems intuitive, it has not been systematically studied. We investigate whether upstream retrieval metrics can serve as reliable early indicators of the final generated response's information coverage. Through experiments across two text RAG benchmarks (TREC NeuCLIR 2024 and TREC RAG 2024) and one multimodal benchmark (WikiVideo), we analyze 15 text retrieval stacks and 10 multimodal retrieval stacks across four RAG pipelines and multiple evaluation frameworks (Auto-ARGUE and MiRAGE). Our findings demonstrate strong correlations between coverage-based retrieval metrics and nugget coverage in generated responses at both topic and system levels. This relationship holds most strongly when retrieval objectives align with generation goals, though more complex iterative RAG pipelines can partially decouple generation quality from retrieval effectiveness. These findings provide empirical support for using retrieval metrics as proxies for RAG performance.

Generative Engine Optimization (GEO) aims to improve content visibility in AI-generated responses. However, existing methods measure contribution-how much a document influences a response-rather than citation, the mechanism that actually drives traffic back to creators. Also, these methods apply generic rewriting rules uniformly, failing to diagnose why individual document are not cited. This paper introduces a diagnostic approach to GEO that asks why a document fails to be cited and intervenes accordingly. We develop a unified framework comprising: (1) the first taxonomy of citation failure modes spanning different stages of a citation pipeline; (2) AgentGEO, an agentic system that diagnoses failures using this taxonomy, selects targeted repairs from a corresponding tool library, and iterates until citation is achieved; and (3) a document-centric benchmark evaluating whether optimizations generalize across held-out queries. AgentGEO achieves over 40% relative improvement in citation rates while modifying only 5% of content, compared to 25% for baselines. Our analysis reveals that generic optimization can harm long-tail content and some documents face challenges that optimization alone cannot fully address-findings with implications for equitable visibility in AI-mediated information access.

Large Language Model (LLM) has exhibited strong reasoning ability in text-based contexts across various domains, yet the limitation of context window poses challenges for the model on long-range inference tasks and necessitates a memory storage system. While many current storage approaches have been proposed with episodic notes and graph representations of memory, retrieval methods still primarily rely on predefined workflows or static similarity top-k over embeddings. To address this inflexibility, we introduced a novel tool-augmented autonomous memory retrieval framework (TA-Mem), which contains: (1) a memory extraction LLM agent which is prompted to adaptively chuck an input into sub-context based on semantic correlation, and extract information into structured notes, (2) a multi-indexed memory database designed for different types of query methods including both key-based lookup and similarity-based retrieval, (3) a tool-augmented memory retrieval agent which explores the memory autonomously by selecting appropriate tools provided by the database based on the user input, and decides whether to proceed to the next iteration or finalizing the response after reasoning on the fetched memories. The TA-Mem is evaluated on the LoCoMo dataset, achieving significant performance improvements over existing baseline approaches. In addition, an analysis of tool use across different question types also demonstrates the adaptivity of the proposed method.

Large-scale scientific collaborations, such as the Compact Muon Solenoid (CMS) at CERN, produce a vast and ever-growing corpus of internal documentation. Navigating this complex information landscape presents a significant challenge for both new and experienced researchers, hindering knowledge sharing and slowing down the pace of scientific discovery. To address this, we present a prototype of MITRA, a Retrieval-Augmented Generation (RAG) based system, designed to answer specific, context-aware questions about physics analyses. MITRA employs a novel, automated pipeline using Selenium for document retrieval from internal databases and Optical Character Recognition (OCR) with layout parsing for high-fidelity text extraction. Crucially, MITRA's entire framework, from the embedding model to the Large Language Model (LLM), is hosted on-premise, ensuring that sensitive collaboration data remains private. We introduce a two-tiered vector database architecture that first identifies the relevant analysis from abstracts before focusing on the full documentation, resolving potential ambiguities between different analyses. We demonstrate the prototype's superior retrieval performance against a standard keyword-based baseline on realistic queries and discuss future work towards developing a comprehensive research agent for large experimental collaborations.

Many modern retrieval problems are set-valued: given a broad intent, the system must return a collection of results that optimizes higher-order properties (e.g., diversity, coverage, complementarity, coherence) while remaining grounded with respect to a fixed database. Set-valued objectives are typically non-decomposable and are not captured by existing supervised (query, content) datasets which only prioritize top-1 retrieval. Consequently, fan-out retrieval is often employed to generate diverse subqueries to retrieve item sets. While reinforcement learning (RL) can optimize set-level objectives via interaction, deploying an RL-tuned LLM for fan-out retrieval is prohibitively expensive at inference time. Conversely, diffusion-based generative retrieval enables efficient single-pass fan-out in embedding space, but requires objective-aligned training targets. To address these issues, we propose R4T (Retrieve-for-Train), which uses RL once as an objective transducer in a three-step process: (i) train a fan-out LLM with composite set-level rewards, (ii) synthesize objective-consistent training pairs, and (iii) train a lightweight diffusion retriever to model the conditional distribution of set-valued outputs. Across large-scale fashion and music benchmarks consisting of curated item sets, we show that R4T improves retrieval quality relative to strong baselines while reducing query-time fan-out latency by an order of magnitude.

Information retrieval (IR) benchmarks typically follow the Cranfield paradigm, relying on static and predefined corpora. However, temporal changes in technical corpora, such as API deprecations and code reorganizations, can render existing benchmarks stale. In our work, we investigate how temporal corpus drift affects FreshStack, a retrieval benchmark focused on technical domains. We examine two independent corpus snapshots of FreshStack from October 2024 and October 2025 to answer questions about LangChain. Our analysis shows that all but one query posed in 2024 remain fully supported by the 2025 corpus, as relevant documents "migrate" from LangChain to competitor repositories, such as LlamaIndex. Next, we compare the accuracy of retrieval models on both snapshots and observe only minor shifts in model rankings, with overall strong correlation of up to 0.978 Kendall $τ$ at Recall@50. These results suggest that retrieval benchmarks re-judged with evolving temporal corpora can remain reliable for retrieval evaluation. We publicly release all our artifacts at https://github.com/fresh-stack/driftbench.

Large Language Model (LLM) agents can automate data-science workflows, but many rigorous statistical methods implemented in R remain underused because LLMs struggle with statistical knowledge and tool retrieval. Existing retrieval-augmented approaches focus on function-level semantics and ignore data distribution, producing suboptimal matches. We propose DARE (Distribution-Aware Retrieval Embedding), a lightweight, plug-and-play retrieval model that incorporates data distribution information into function representations for R package retrieval. Our main contributions are: (i) RPKB, a curated R Package Knowledge Base derived from 8,191 high-quality CRAN packages; (ii) DARE, an embedding model that fuses distributional features with function metadata to improve retrieval relevance; and (iii) RCodingAgent, an R-oriented LLM agent for reliable R code generation and a suite of statistical analysis tasks for systematically evaluating LLM agents in realistic analytical scenarios. Empirically, DARE achieves an NDCG at 10 of 93.47%, outperforming state-of-the-art open-source embedding models by up to 17% on package retrieval while using substantially fewer parameters. Integrating DARE into RCodingAgent yields significant gains on downstream analysis tasks. This work helps narrow the gap between LLM automation and the mature R statistical ecosystem.

Sequential Recommendation (SR) predicts users next interactions by modeling the temporal order of their historical behaviors. Existing approaches, including traditional sequential models and generative recommenders, achieve strong performance but primarily rely on explicit interactions such as clicks or purchases while overlooking item exposures. This ignorance introduces selection bias, where exposed but unclicked items are misinterpreted as disinterest, and exposure bias, where unexposed items are treated as irrelevant. Effectively addressing these biases requires distinguishing between items that were "not exposed" and those that were "not of interest", which cannot be reliably inferred from correlations in historical data. Counterfactual reasoning provides a natural solution by estimating user preferences under hypothetical exposure, and Inverse Propensity Scoring (IPS) is a common tool for such estimation. However, conventional IPS methods are static and fail to capture the sequential dependencies and temporal dynamics of user behavior. To overcome these limitations, we propose Time aware Inverse Propensity Scoring (TIPS). Unlike traditional static IPS, TIPS effectively accounts for sequential dependencies and temporal dynamics, thereby capturing user preferences more accurately. Extensive experiments show that TIPS consistently enhances recommendation performance as a plug-in for various sequential recommenders. Our code will be publicly available upon acceptance.

Retrieval-Augmented Generation (RAG) enhances large language models by incorporating external knowledge. However, existing vector-based methods often fail on global sensemaking tasks that require reasoning across many documents. GraphRAG addresses this by organizing documents into a knowledge graph with hierarchical communities that can be recursively summarized. Current GraphRAG approaches rely on Leiden clustering for community detection, but we prove that on sparse knowledge graphs, where average degree is constant and most nodes have low degree, modularity optimization admits exponentially many near-optimal partitions, making Leiden-based communities inherently non-reproducible. To address this, we propose replacing Leiden with k-core decomposition, which yields a deterministic, density-aware hierarchy in linear time. We introduce a set of lightweight heuristics that leverage the k-core hierarchy to construct size-bounded, connectivity-preserving communities for retrieval and summarization, along with a token-budget-aware sampling strategy that reduces LLM costs. We evaluate our methods on real-world datasets including financial earnings transcripts, news articles, and podcasts, using three LLMs for answer generation and five independent LLM judges for head-to-head evaluation. Across datasets and models, our approach consistently improves answer comprehensiveness and diversity while reducing token usage, demonstrating that k-core-based GraphRAG is an effective and efficient framework for global sensemaking.

Most large-scale recommender systems follow a multi-stage cascade of retrieval, pre-ranking, ranking, and re-ranking. A key challenge at the pre-ranking stage arises from the heterogeneity of training instances sampled from coarse-grained retrieval results, fine-grained ranking signals, and exposure feedback. Our analysis reveals that prevailing pre-ranking methods, which indiscriminately mix heterogeneous samples, suffer from gradient conflicts: hard samples dominate training while easy ones remain underutilized, leading to suboptimal performance. We further show that the common practice of uniformly scaling model complexity across all samples is inefficient, as it overspends computation on easy cases and slows training without proportional gains. To address these limitations, this paper presents Heterogeneity-Aware Adaptive Pre-ranking (HAP), a unified framework that mitigates gradient conflicts through conflict-sensitive sampling coupled with tailored loss design, while adaptively allocating computational budgets across candidates. Specifically, HAP disentangles easy and hard samples, directing each subset along dedicated optimization paths. Building on this separation, it first applies lightweight models to all candidates for efficient coverage, and further engages stronger models on the hard ones, maintaining accuracy while reducing cost. This approach not only improves pre-ranking effectiveness but also provides a practical perspective on scaling strategies in industrial recommender systems. HAP has been deployed in the Toutiao production system for 9 months, yielding up to 0.4% improvement in user app usage duration and 0.05% in active days, without additional computational cost. We also release a large-scale industrial hybrid-sample dataset to enable the systematic study of source-driven candidate heterogeneity in pre-ranking.

Shared-account usage is common on streaming and e-commerce platforms, where multiple users share one account. Existing shared-account sequential recommendation (SSR) methods often assume a fixed number of latent users per account, limiting their ability to adapt to diverse sharing patterns and reducing recommendation accuracy. Recent latent reasoning technique applied in sequential recommendation (SR) generate intermediate embeddings from the user embedding (e.g, last item embedding) to uncover users' potential interests, which inspires us to treat the problem of inferring the number of latent users as generating a series of intermediate embeddings, shifting from inferring preferences behind user to inferring the users behind account. However, the last item cannot be directly used for reasoning in SSR, as it can only represent the behavior of the most recent latent user, rather than the collective behavior of the entire account. To address this, we propose DisenReason, a two-stage reasoning method tailored to SSR. DisenReason combines behavior disentanglement stage from frequency-domain perspective to create a collective and unified account behavior representation, which serves as a pivot for latent user reasoning stage to infer the number of users behind the account. Experiments on four benchmark datasets show that DisenReason consistently outperforms all state-of-the-art baselines across four benchmark datasets, achieving relative improvements of up to 12.56\% in MRR@5 and 6.06\% in Recall@20.

Attention mechanism remains the defining operator in Transformers since it provides expressive global credit assignment, yet its $O(N^2 d)$ time and memory cost in sequence length $N$ makes long-context modeling expensive and often forces truncation or other heuristics. Linear attention reduces complexity to $O(N d^2)$ by reordering computation through kernel feature maps, but this reformulation drops the softmax mechanism and shifts the attention score distribution. In recommender systems, low-rank structure in matrices is not a rare case, but rather the default inductive bias in its representation learning, particularly explicit in the user behavior sequence modeling. Leveraging this structure, we introduce SVD-Attention, which is theoretically lossless on low-rank matrices and preserves softmax while reducing attention complexity from $O(N^2 d)$ to $O(Ndr)$. With SVD-Attention, we propose SOLAR, SVD-Optimized Lifelong Attention for Recommendation, a sequence modeling framework that supports behavior sequences of ten-thousand scale and candidate sets of several thousand items in cascading process without any filtering. In Kuaishou's online recommendation scenario, SOLAR delivers a 0.68\% Video Views gain together with additional business metrics improvements.

The Generator-Evaluator (G-E) framework, i.e., evaluating K sequences from a generator and selecting the top-ranked one according to evaluator scores, is a foundational paradigm in tasks such as Recommender Systems (RecSys) and Natural Language Processing (NLP). Traditional evaluators process sequences independently, suffering from two major limitations: (1) lack of explicit cross-sequence comparison, leading to suboptimal accuracy; (2) poor parallelization with linear complexity of O(K), resulting in inefficient resource utilization and negative impact on both throughput and latency. To address these challenges, we propose FlashEvaluator, which enables cross-sequence token information sharing and processes all sequences in a single forward pass. This yields sublinear computational complexity that improves the system's efficiency and supports direct inter-sequence comparisons that improve selection accuracy. The paper also provides theoretical proofs and extensive experiments on recommendation and NLP tasks, demonstrating clear advantages over conventional methods. Notably, FlashEvaluator has been deployed in online recommender system of Kuaishou, delivering substantial and sustained revenue gains in practice.

Can we design effective recommender systems free from users, IDs, and GNNs? Recommender systems are central to personalized content delivery across domains, with top-K item recommendation being a fundamental task to retrieve the most relevant items from historical interactions. Existing methods rely on entrenched design conventions, often adopted without reconsideration, such as storing per-user embeddings (user-dependent), initializing features from raw IDs (ID-dependent), and employing graph neural networks (GNN-dependent). These dependencies incur several limitations, including high memory costs, cold-start and over-smoothing issues, and poor generalization to unseen interactions. In this work, we propose AlphaFree, a novel recommendation method free from users, IDs, and GNNs. Our main ideas are to infer preferences on-the-fly without user embeddings (user-free), replace raw IDs with language representations (LRs) from pre-trained language models (ID-free), and capture collaborative signals through augmentation with similar items and contrastive learning, without GNNs (GNN-free). Extensive experiments on various real-world datasets show that AlphaFree consistently outperforms its competitors, achieving up to around 40% improvements over non-LR-based methods and up to 5.7% improvements over LR-based methods, while significantly reducing GPU memory usage by up to 69% under high-dimensional LRs.

Item-side fairness is crucial for ensuring the fair exposure of long-tail items in interactive recommender systems. Existing approaches promote the exposure of long-tail items by directly incorporating them into recommended results. This causes misalignment between user preferences and the recommended long-tail items, which hinders long-term user engagement and reduces the effectiveness of recommendations. We aim for a proactive fairness-guiding strategy, which actively guides user preferences toward long-tail items while preserving user satisfaction during the interactive recommendation process. To this end, we propose HRL4PFG, an interactive recommendation framework that leverages hierarchical reinforcement learning to guide user preferences toward long-tail items progressively. HRL4PFG operates through a macro-level process that generates fairness-guided targets based on multi-step feedback, and a micro-level process that fine-tunes recommendations in real time according to both these targets and evolving user preferences. Extensive experiments show that HRL4PFG improves cumulative interaction rewards and maximum user interaction length by a larger margin when compared with state-of-the-art methods in interactive recommendation environments.

Retrieval-Augmented Generation (RAG) grounds Large Language Models (LLMs) to mitigate factual hallucinations. Recent paradigms shift from static pipelines to Modular and Agentic RAG frameworks, granting models autonomy for multi-hop reasoning or self-correction. However, current reflective RAG heavily relies on massive LLMs as universal evaluators. In high-throughput systems, executing complete forward passes for billion-parameter models merely for binary routing introduces severe computational redundancy. Furthermore, in autonomous agent scenarios, inaccurate retrieval causes models to expend excessive tokens on spurious reasoning and redundant tool calls, inflating Time-to-First-Token (TTFT) and costs. We propose Tiny-Critic RAG, decoupling evaluation by deploying a parameter-efficient Small Language Model (SLM) via Low-Rank Adaptation (LoRA). Acting as a deterministic gatekeeper, Tiny-Critic employs constrained decoding and non-thinking inference modes for ultra-low latency binary routing. Evaluations on noise-injected datasets demonstrate Tiny-Critic RAG achieves routing accuracy comparable to GPT-4o-mini while reducing latency by an order of magnitude, establishing a highly cost-effective paradigm for agent deployment.

Complex clinical decision making often fails not because a model lacks facts, but because it cannot reliably select and apply the right procedural knowledge and the right prior example at the right reasoning step. We frame clinical question answering as an agent problem with two explicit, retrievable resources: skills, reusable clinical procedures such as guidelines, protocols, and pharmacologic mechanisms; and experience, verified reasoning trajectories from previously solved cases (e.g., chain-of-thought solutions and their step-level decompositions). At test time, the agent retrieves both relevant skills and experiences from curated libraries and performs lightweight test-time adaptation to align the language model's intermediate reasoning with clinically valid logic. Concretely, we build (i) a skills library from guideline-style documents organized as executable decision rules, (ii) an experience library of exemplar clinical reasoning chains indexed by step-level transitions, and (iii) a step-aware retriever that selects the most useful skill and experience items for the current case. We then adapt the model on the retrieved items to reduce instance-step misalignment and to prevent reasoning from drifting toward unsupported shortcuts. Experiments on medical question-answering benchmarks show consistent gains over strong medical RAG baselines and prompting-only reasoning methods. Our results suggest that explicitly separating and retrieving clinical skills and experience, and then aligning the model at test time, is a practical approach to more reliable medical agents.

Multimodal embedding models, rooted in multimodal large language models (MLLMs), have yielded significant performance improvements across diverse tasks such as retrieval and classification. However, most existing approaches rely heavily on large-scale contrastive learning, with limited exploration of how the architectural and training paradigms of MLLMs affect embedding quality. While effective for generation, the causal attention and next-token prediction paradigm of MLLMs does not explicitly encourage the formation of globally compact representations, limiting their effectiveness as multimodal embedding backbones. To address this, we propose CoCoA, a Content reconstruction pre-training paradigm based on Collaborative Attention for multimodal embedding optimization. Specifically, we restructure the attention flow and introduce an EOS-based reconstruction task, encouraging the model to reconstruct input from the corresponding <EOS> embeddings. This drives the multimodal model to compress the semantic information of the input into the <EOS> token, laying the foundations for subsequent contrastive learning. Extensive experiments on MMEB-V1 demonstrate that CoCoA built upon Qwen2-VL and Qwen2.5-VL significantly improves embedding quality. Results validate that content reconstruction serves as an effective strategy to maximize the value of existing data, enabling multimodal embedding models generate compact and informative representations, raising their performance ceiling.

Personal photo albums are not merely collections of static images but living, ecological archives defined by temporal continuity, social entanglement, and rich metadata, which makes the personalized photo retrieval non-trivial. However, existing retrieval benchmarks rely heavily on context-isolated web snapshots, failing to capture the multi-source reasoning required to resolve authentic, intent-driven user queries. To bridge this gap, we introduce PhotoBench, the first benchmark constructed from authentic, personal albums. It is designed to shift the paradigm from visual matching to personalized multi-source intent-driven reasoning. Based on a rigorous multi-source profiling framework, which integrates visual semantics, spatial-temporal metadata, social identity, and temporal events for each image, we synthesize complex intent-driven queries rooted in users' life trajectories. Extensive evaluation on PhotoBench exposes two critical limitations: the modality gap, where unified embedding models collapse on non-visual constraints, and the source fusion paradox, where agentic systems perform poor tool orchestration. These findings indicate that the next frontier in personal multimodal retrieval lies beyond unified embeddings, necessitating robust agentic reasoning systems capable of precise constraint satisfaction and multi-source fusion. Our PhotoBench is available.

Click-through rate (CTR) models in advertising and recommendation systems rely heavily on item ID embeddings, which struggle in item cold-start settings. We present IDProxy, a solution that leverages multimodal large language models (MLLMs) to generate proxy embeddings from rich content signals, enabling effective CTR prediction for new items without usage data. These proxies are explicitly aligned with the existing ID embedding space and are optimized end-to-end under CTR objectives together with the ranking model, allowing seamless integration into existing large-scale ranking pipelines. Offline experiments and online A/B tests demonstrate the effectiveness of IDProxy, which has been successfully deployed in both Content Feed and Display Ads features of Xiaohongshu's Explore Feed, serving hundreds of millions of users daily.

Multimodal retrieval is the task of aggregating information from queries across heterogeneous modalities to retrieve desired targets. State-of-the-art multimodal retrieval models can understand complex queries, yet they are typically limited to two modalities: text and vision. This limitation impedes the development of universal retrieval systems capable of comprehending queries that combine more than two modalities. To advance toward this goal, we present OmniRet, the first retrieval model capable of handling complex, composed queries spanning three key modalities: text, vision, and audio. Our OmniRet model addresses two critical challenges for universal retrieval: computational efficiency and representation fidelity. First, feeding massive token sequences from modality-specific encoders to Large Language Models (LLMs) is computationally inefficient. We therefore introduce an attention-based resampling mechanism to generate compact, fixed-size representations from these sequences. Second, compressing rich omni-modal data into a single embedding vector inevitably causes information loss and discards fine-grained details. We propose Attention Sliced Wasserstein Pooling to preserve these fine-grained details, leading to improved omni-modal representations. OmniRet is trained on an aggregation of approximately 6 million query-target pairs spanning 30 datasets. We benchmark our model on 13 retrieval tasks and a MMEBv2 subset. Our model demonstrates significant improvements on composed query, audio and video retrieval tasks, while achieving on-par performance with state-of-the-art models on others. Furthermore, we curate a new Audio-Centric Multimodal Benchmark (ACM). This new benchmark introduces two critical, previously missing tasks-composed audio retrieval and audio-visual retrieval to more comprehensively evaluate a model's omni-modal embedding capacity.

Retrieval-Augmented Generation (RAG) systems commonly adopt retrieval fusion techniques such as multi-query retrieval and reciprocal rank fusion (RRF) to increase document recall, under the assumption that higher recall leads to better answer quality. While these methods show consistent gains in isolated retrieval benchmarks, their effectiveness under realistic production constraints remains underexplored. In this work, we evaluate retrieval fusion in a production-style RAG pipeline operating over an enterprise knowledge base, with fixed retrieval depth, re-ranking budgets, and latency constraints. Across multiple fusion configurations, we find that retrieval fusion does increase raw recall, but these gains are largely neutralized after re-ranking and truncation. In our setting, fusion variants fail to outperform single-query baselines on KB-level Top-$k$ accuracy, with Hit@10 decreasing from $0.51$ to $0.48$ in several configurations. Moreover, fusion introduces additional latency overhead due to query rewriting and larger candidate sets, without corresponding improvements in downstream effectiveness. Our analysis suggests that recall-oriented fusion techniques exhibit diminishing returns once realistic re-ranking limits and context budgets are applied. We conclude that retrieval-level improvements do not reliably translate into end-to-end gains in production RAG systems, and argue for evaluation frameworks that jointly consider retrieval quality, system efficiency, and downstream impact.

Many readers today struggle to assess the trustworthiness of online news because reliable reporting coexists with misinformation. The TREC 2025 DRAGUN (Detection, Retrieval, and Augmented Generation for Understanding News) Track provided a venue for researchers to develop and evaluate assistive RAG systems that support readers' news trustworthiness assessment by producing reader-oriented, well-attributed reports. As the organizers of the DRAGUN track, we describe the resources that we have newly developed to allow for the reuse of the track's tasks. The track had two tasks: (Task 1) Question Generation, producing 10 ranked investigative questions; and (Task 2, the main task) Report Generation, producing a 250-word report grounded in the MS MARCO V2.1 Segmented Corpus. As part of the track's evaluation, we had TREC assessors create importance-weighted rubrics of questions with expected short answers for 30 different news articles. These rubrics represent the information that assessors believe is important for readers to assess an article's trustworthiness. The assessors then used their rubrics to manually judge the participating teams' submitted runs. To make these tasks and their rubrics reusable, we have created an automated process to judge runs not part of the original assessing. We show that our AutoJudge ranks existing runs well compared to the TREC human-assessed evaluation (Kendall's $τ= 0.678$ for Task 1 and $τ= 0.872$ for Task 2). These resources enable both the evaluation of RAG systems for assistive news trustworthiness assessment and, with the human evaluation as a benchmark, research on improving automated RAG evaluation.

We present Geodesic Semantic Search (GSS), a retrieval system that learns node-specific Riemannian metrics on citation graphs to enable geometry-aware semantic search. Unlike standard embedding-based retrieval that relies on fixed Euclidean distances, \gss{} learns a low-rank metric tensor $\mL_i \in \R^{d \times r}$ at each node, inducing a local positive semi-definite metric $\mG_i = \mL_i \mL_i^\top + \eps \mI$. This parameterization guarantees valid metrics while keeping the model tractable. Retrieval proceeds via multi-source Dijkstra on the learned geodesic distances, followed by Maximal Marginal Relevance reranking and path coherence filtering. On citation prediction benchmarks with 169K papers, \gss{} achieves 23\% relative improvement in Recall@20 over SPECTER+FAISS baselines while providing interpretable citation paths. Our hierarchical coarse-to-fine search with k-means pooling reduces computational cost by 4$\times$ compared to flat geodesic search while maintaining 97\% retrieval quality. We provide theoretical analysis of when geodesic distances outperform direct similarity, characterize the approximation quality of low-rank metrics, and validate predictions empirically. Code and trained models are available at https://github.com/YCRG-Labs/geodesic-search.

Agentic search has emerged as a promising paradigm for adaptive retrieval systems powered by large language models (LLMs). However, existing benchmarks primarily focus on quality, overlooking efficiency factors that are critical for real-world deployment. Moreover, real-world user queries often contain underspecified preferences, a challenge that remains largely underexplored in current agentic search evaluation. As a result, many agentic search systems remain impractical despite their impressive performance. In this work, we introduce HotelQuEST, a benchmark comprising 214 hotel search queries that range from simple factual requests to complex queries, enabling evaluation across the full spectrum of query difficulty. We further address the challenge of evaluating underspecified user preferences by collecting clarifications that make annotators' implicit preferences explicit for evaluation. We find that LLM-based agents achieve higher accuracy than traditional retrievers, but at substantially higher costs due to redundant tool calls and suboptimal routing that fails to match query complexity to model capability. Our analysis exposes inefficiencies in current agentic search systems and demonstrates substantial potential for cost-aware optimization.