The Inference Report

Decentralization as a concept in computer science has existed for over half a century. Despite its fundamental role across domains such as security, distributed computing, artificial intelligence, cloud infrastructures, and Internet of Things (IoT) architectures, there remains no universally accepted definition of decentralization applicable across computer communication systems. This has become increasingly problematic with the emergence of decentralized AI and machine learning paradigms, including collaborative training, distributed inference, blockchain-based, and agentic AI, where decentralization is often treated as a core design objective. Meanwhile, existing approaches frequently conflate decentralization with related notions such as distribution of trust or specific implementation paradigms. Such ambiguity creates inconsistencies in system analysis, limits comparability between works, and weakens the rigor of formal reasoning surrounding communication architectures and protocol design. In this work, we define this research gap as the Decentralization Problem. We analyze the formal-semantic, epistemological, and pragmatic foundations of decentralization and introduce a graph-based ontology defining it as both relational and subject-specific property of computer communication systems. The framework formally distinguishes decentralization from distribution and supports evaluation through two novel metrics: Void Tolerance and Imperviousness. We also provide a browser-based implementation that enables automated classification and metric computation of arbitrary systems. Instantiations to federated learning and blockchain architectures show consistent, comparable assessments where existing definitions produce incomplete or contradictory conclusions, providing a domain-independent foundation for analysing decentralization across heterogeneous systems.

Flexible macromolecular docking offers high-fidelity predictions of biomolecular interactions, but remains prohibitively expensive at scale. Among existing approaches, LightDock leverages Glowworm Swarm Optimization (GSO) for accuracy, yet suffers from limited parallelism, irregular computation, and severe load imbalance, preventing efficient execution on GPU supercomputers. We present SparkleDock, a scalable GSO-based docking framework enabling near-real-time flexible docking. We redesign GSO to expose massive fine-grained parallelism at the glowworm-agent level, and restructure the dominant energy scoring computation into a Tensor Core-compatible formulation, enabling efficient execution of irregular pairwise interactions through structured matrix operations. We further introduce a performance-model-driven scheduling for load balancing and out-of-core scaling across GPUs. SparkleDock achieves 9.7 $\times$ and 18.9 $\times$ speedups over LightDock on single A100 and H100 GPU, and delivers over two orders of magnitude acceleration at scale. On 512 GPUs, it reduces docking time from hours to seconds, enabling large-scale, high-fidelity virtual screening previously impractical with flexible docking.

Agentic science is transforming the landscape of computational work, extending to scientific pipelines and workload managers. The workloads require specialized hardware within and across institutions. If assessing workload needs against environments is required for scheduling, automated discovery of resources is an essential step. In this paper, we present a hierarchical, dynamic architecture and software to discover resources across diverse cloud, edge, and HPC systems. The design enables concurrent, asynchronous negotiation, selection, and dispatch of requests for work using secretary agents. The agents probe and discover 51 real and simulated providers across 7 categories. We perform 19,973 negotiation and 6,952 selection simulations to assess reliability of decisions, demonstrating high (87.71\%) negotiation accuracy and selection costs comparable to more traditional strategies. Designed for extensibility and currently supporting the Genesis Mission, this architecture exemplifies the importance of careful coordination between agents, discovery tools, and infrastructure for agentic science.

We report operational experience full-fine-tuning a 32.76B-parameter dense model (Qwen3-32B) on 16 x NVIDIA B300 (two nodes, FSDP / ZeRO-3) -- among the first published field accounts on this accelerator. We claim no new algorithm. The individual mechanisms we use are established practice; our contribution is the integrated field experience and a set of calibrated measurements on new hardware. Concretely we offer four practitioner artifacts. (1) A B300-calibrated power-draw triage table that distinguishes compute / communication / data-starvation / checkpoint-or-deadlock / idle by board wattage (utilization% reads 100% during an NCCL hang). (2) A set of honest negative results that dispel common optimization folklore at this scale: a controlled A/B in which per-step NFS reading matches a pretokenized local cache (~53k tok/s) because the corpus fits in page cache and the job is compute-bound; and a reconstruction of an earlier "throughput collapse" as NFS/CPU contention rather than a storage-medium limit. (3) Calibrated 4/8/16-GPU strong-scaling and GPU-hour numbers on B300 (near-linear, as expected in this regime; we report absolute values as reference data). (4) A worked failure case -- an epoch-end NCCL deadlock from per-rank token-packing imbalance -- together with a 2.7-second pre-run invariant gate and an external watcher that turn multi-hour silent failures into instant rejections. This deadlock and its remedy correspond to PyTorch's documented Join / equalize-to-minimum practice; we position our instantiation against that prior art and report the GPU-hours the failure cost and the gate saves. The transferable takeaway is operational, not algorithmic: for data-dependent data-parallel jobs, watch power rather than utilization, and verify invariants before launch -- a passing smoke test is not evidence of a safe full run.

Attention--Feed-Forward Network (FFN) Disaggregation (AFD) is emerging as a promising architecture for serving Mixture-of-Experts (MoE) language models. While existing AFD systems improve the efficiency of disaggregated execution, they leave a deployment question unanswered: under the same model, workload, time-per-output-token (TPOT) service-level objective (SLO), hardware budget, hardware catalog, and runtime capabilities, does AFD provide higher throughput than the best collocated deployment? Answering this question requires jointly optimizing hardware assignment and deployment organization for both architectures, making exhaustive provisioning prohibitively expensive. We present AFD-Ledger, an offline analytical provisioning system that independently provisions AFD and collocated deployments using an analytical execution model and an evaluation-bounded hardware search. Across deployment spaces where exhaustive provisioning is feasible, AFD-Ledger reduces complete deployment evaluations by 68.8%--83.5% while still recovering the globally optimal deployment. On three physical LongCat 2.0 deployments, it preserves the correct architecture decision while predicting AFD-to-collocated throughput within 6.6%--9.6% of measurement. Using this validated framework, we show that homogeneous AFD improves fixed-budget throughput in only a minority of the studied settings, heterogeneous AFD requires deployment-level hardware complementarity rather than heuristic device selection, and role-specific hardware improvements matter primarily when they enable better deployment organizations by crossing deployment capability--price boundaries.

Sparse matrix kernels are fundamental to scientific computing, graph analytics, and machine learning. Their GPU performance depends strongly on the input sparsity pattern and execution strategy. For the same SpMM on the same matrix, cuSPARSE exhibits a 350x performance gap between CSR and Blocked-ELL. Our study of multiple data formats, specialized systems, and sparse compilers shows that no single implementation consistently dominates across sparsity patterns and operators. This motivates a system that can adapt its representation, execution strategy, and hardware mapping to each workload and target GPU. We present SparseDitto, an LLM-based system that constructs a GPU kernel for each matrix, operator, and target GPU. SparseDitto supports SpMV, SpMM, and SpGEMM within a unified design framework. A lightweight additive model ranks established strategies using structural features of the input matrix. An architecture-aware planner then proposes several candidate designs. Coding and verification agents implement and refine them using measurements from the target GPU. Across three sparse operators and a diverse set of matrices, SparseDitto achieves a geometric-mean speedup of 2.68x over cuSPARSE on an NVIDIA RTX PRO 6000 GPU, with a maximum of 146.61x. On an NVIDIA H200 GPU, it achieves 2.79x, with a maximum of 78.5x. Its generated SpMM kernels also accelerate full-batch GCN training by up to 3.39x.