Clay weekly context brief for the Computer Science category (ISO week 2026-W43).

Clay tracks publications from the Computer Science feed list. Below are recent items from this category, each with its source and a short description of what the publication covers when one is available in the source feed.

Recent publications:
1. Evaluating Model Capabilities After Benchmark Saturation
   Source: cs.AI (Artificial Intelligence)
   Link: https://arxiv.org/abs/2610.20667
   Rapid advances in AI capabilities are causing benchmarks to saturate.
2. Helix: A Unified Programmable Accelerator for Diverse Krylov Subspace Solvers
   Source: cs.AR (Hardware Architecture)
   Link: https://arxiv.org/abs/2610.21096
   Domain-specific accelerators deliver high performance and energy efficiency through specialization but often lack support for extensible algorithm families.
3. Restless Guards: Robust Guarding of Polyominoes
   Source: cs.CG (Computational Geometry)
   Link: https://arxiv.org/abs/2610.21793
   The art gallery problem is a fundamental problem whose practical applications have inspired numerous variants.
4. PRISM: PRIvileged Sibling-guided Multi-Agent Topology Design
   Source: cs.CL (Computation and Language)
   Link: https://arxiv.org/abs/2610.20628
   Multi-agent systems (MAS) based on large language models (LLMs) have become a powerful way to address complex problems, and their effectiveness depends critically on the collaboration topology that decides which agents participate and how they communicate.
5. Unmasking the Secret Code: Using Canaries to Show How Masked Context Controls Memorization in SFT
   Source: cs.CR (Cryptography and Security)
   Link: https://arxiv.org/abs/2610.20656
   Canaries, or synthetic training records containing known secrets, are widely used to assess whether language models reveal information from their training data.
6. Towards automatic digital garment assembly
   Source: cs.CV (Computer Vision and Pattern Recognition)
   Link: https://arxiv.org/abs/2610.20651
   Creating digital garments from existing two-dimensional patterns remains a time-consuming process.
7. Learning to Transfer School Attendance Zoning across Cities
   Source: cs.CY (Computers and Society)
   Link: https://arxiv.org/abs/2610.21924
   Designing multi-school attendance zones requires selecting constraint settings that balance educational equity, accessibility, and school capacity.
8. Reflective Compositional Grounding for Natural-Language-to-SPARQL
   Source: cs.DB (Databases)
   Link: https://arxiv.org/abs/2610.21796
   Natural-language query generation over large knowledge graphs embodies the classical vocabulary problem in human-systems communication.
9. Probabilistic Optical-Layer Surrogate for Risk-Aware Control-Plane Decision Making
   Source: cs.DC (Distributed, Parallel, and Cluster Computing)
   Link: https://arxiv.org/abs/2610.21069
   In this paper, we propose a generative surrogate for the optical layer.
10. ResStreamGS: Residual-Sensitive Dual-Stream Gaussian Splatting for Mobile Volumetric Video Streaming
   Source: cs.GR (Graphics)
   Link: https://arxiv.org/abs/2610.21121
   Real-time mobile volumetric video enables immersive free-viewpoint experiences, but delivering high-quality 3D content under constrained bandwidth and computation remains challenging.
11. Faster alone vs. further together: The speed and impact of researchers' responses to societal crises
   Source: cs.DL (Digital Libraries)
   Link: https://arxiv.org/abs/2603.19246
   Throughout the 21st century, society encountered numerous crises that demanded quick and widespread action.
12. The Koml\'os Problem in Matrix-Multiplication Time
   Source: cs.DM (Discrete Mathematics)
   Link: https://arxiv.org/abs/2610.20995
   We give a randomized algorithm that, for an $m\times n$ matrix whose columns have Euclidean norm at most one, finds a signing with discrepancy below $99$ in expected $\widetilde O(L_{\mathrm{in}}+n^\omega)$ bit operations, where $L_{\mathrm{in}}$ is the input encoding length and $\omega$ is the exponent of matrix multiplication.
13. Eccentricity Shortest Path: the structural parameter landscape
   Source: cs.DS (Data Structures and Algorithms)
   Link: https://arxiv.org/abs/2610.20769
   The Eccentricity Shortest Path problem asks, given a connected graph and a nonnegative integer $R$, whether some shortest path is within distance $R$ of every vertex.
14. Reducing QUBO Complexity in Drone Delivery Through Classical Feasibility Preprocessing
   Source: cs.ET (Emerging Technologies)
   Link: https://arxiv.org/abs/2610.21003
   Drone-delivery planning combines routing decisions with operational constraints such as payload capacity and limited onboard energy, resulting in combinatorial optimization problems that can become costly to encode directly in binary form.
15. Aligning LLMs with Sparse Retrieval via Adaptive Sparse Supervision
   Source: cs.IR (Information Retrieval)
   Link: https://arxiv.org/abs/2610.21877
   While large language models (LLMs) offer strong semantic representations, their application to first-stage retrieval incurs substantial latency and storage costs.
16. Approximating the Value of Time-Bounded Weighted Timed Games
   Source: cs.GT (Computer Science and Game Theory)
   Link: https://arxiv.org/abs/2610.21487
   The Value Problem for weighted timed games, an extension of timed automata with weights and adversity, consists in determining, given a weighted timed game with a reachability objective and a rational threshold, whether or not the value of the game exceeds the threshold.
17. Same Answer, New Box: How Much of RLVR's Out-of-Domain Gain Is Answer Format?
   Source: cs.LG (Machine Learning)
   Link: https://arxiv.org/abs/2610.20639
   Reinforcement learning with verifiable rewards (RLVR) is credited with gains beyond the domain its rewards check, such as higher MMLU accuracy after training on mathematics.
18. Exploring Mixed Reality for Engagement in Robotics: A Comparison of Robot Control Methods
   Source: cs.HC (Human-Computer Interaction)
   Link: https://arxiv.org/abs/2610.16186
   This paper presents a Mixed Reality (MR) robotics experience that provides an accessible introduction to interacting with a physical robot.
19. Multimodal Remote Inference
   Source: cs.NI (Networking and Internet Architecture)
   Link: https://arxiv.org/abs/2508.07555
   We consider a remote inference system with multiple modalities, where a multimodal machine learning (ML) model performs real-time inference using features collected from remote sensors.
20. Approaching Backpropagation Accuracy with Weight Perturbation on MNIST
   Source: cs.NE (Neural and Evolutionary Computing)
   Link: https://arxiv.org/abs/2610.21665
   Weight perturbation (WP) estimates learning directions from random parameter perturbations and a scalar change in loss, without backpropagating derivatives.
21. Video-Time Priors for EEG-fNIRS Emotion Regression on Familiar Videos
   Source: cs.MM (Multimedia)
   Link: https://arxiv.org/abs/2610.03618
   Do viewer-specific physiological signals improve continuous emotion regression beyond the stimulus-driven response shared across viewers?
22. The Delegation Dilemma: Economies of Scale, Accountability, and Polycentric Governance
   Source: cs.MA (Multiagent Systems)
   Link: https://arxiv.org/abs/2610.21746
   Many collective-action systems are neither fully decentralized nor fully centralized: agents voluntarily organize into institutions that share organizational costs and delegate decisions to representatives.
23. Multiscale Generalized Sparse Plane Wave Decomposition in the Spherical Harmonic Domain
   Source: cs.SD (Sound)
   Link: https://arxiv.org/abs/2610.21425
   Sparse plane wave decomposition (SPWD) is a physics-driven model-based approach for analyzing sound fields that is suitable for real-time operation.
24. Manifold Generative Annealing for Contact Constraint-aware Impedance Control with Reinforcement Learning Prior
   Source: cs.RO (Robotics)
   Link: https://arxiv.org/abs/2610.21140
   Contact-rich robotic tasks require coordinated motion and impedance under uncertain interaction dynamics and physical constraints.
25. DataKernelBench: Can LLMs Optimize Database Queries on GPUs?
   Source: cs.PL (Programming Languages)
   Link: https://arxiv.org/abs/2608.25061
   GPUs increasingly accelerate database systems, but query-specific peak performance still often relies on hand-written kernels.
26. FAME: An FPGA-Based Platform for Approximate Multipliers Evaluation with Pattern-Guided DNN Retraining
   Source: cs.PF (Performance)
   Link: https://arxiv.org/abs/2609.17730
   Approximate multipliers can reduce hardware area and energy consumption in Deep Neural Network (DNN) inference; however, they introduce computational errors.
27. Causal Syntax Circuits in CodeLMs: Budget and Probe Decide What Looks Localised
   Source: cs.SE (Software Engineering)
   Link: https://arxiv.org/abs/2610.20683
   Code models write Python when the prompt does not say which language to use.
28. Single-stage joint diffusion of sky images and clear-sky index for ultra-short-term irradiance forecasting: what the frame-to-irradiance stage discards
   Source: cs.SY (Systems and Control)
   Link: https://arxiv.org/abs/2610.21795
   Sky-image irradiance forecasting is dominated by a two-stage design: a video-prediction model forecasts future sky frames, and a separately trained regressor maps those frames to irradiance.
29. Sparse Contextual Coupling Reshapes Diffusion Geometry in Multilayer Hypergraphs
   Source: cs.SI (Social and Information Networks)
   Link: https://arxiv.org/abs/2605.20454
   Many complex systems combine dense background structure with sparse contextual information.
30. Value-Informed Schedule Trimming for Accelerated Sequential Monte Carlo Guidance in Discrete Diffusion
   Source: cs.AI (Artificial Intelligence)
   Link: https://arxiv.org/abs/2610.20669
   Reward-guided discrete diffusion enables test-time alignment with downstream objectives, but applying Sequential Monte Carlo (SMC) guidance at every denoising step is computationally expensive.
31. PSA: A Pruning-Friendly Sparse-Aware Accelerator for Efficient Sparse Convolution in SNNs
   Source: cs.AR (Hardware Architecture)
   Link: https://arxiv.org/abs/2610.21313
   Spiking neural networks (SNNs) are promising for energy-efficient artificial intelligence due to their event-driven sparse computation, and neuron pruning can further reduce redundant operations.

Sources in this brief: cs.AI (Artificial Intelligence); cs.AR (Hardware Architecture); cs.CG (Computational Geometry); cs.CL (Computation and Language); cs.CR (Cryptography and Security); cs.CV (Computer Vision and Pattern Recognition); cs.CY (Computers and Society); cs.DB (Databases); cs.DC (Distributed, Parallel, and Cluster Computing); cs.DL (Digital Libraries); cs.DM (Discrete Mathematics); cs.DS (Data Structures and Algorithms); cs.ET (Emerging Technologies); cs.GR (Graphics); cs.GT (Computer Science and Game Theory); cs.HC (Human-Computer Interaction); cs.IR (Information Retrieval); cs.LG (Machine Learning); cs.MA (Multiagent Systems); cs.MM (Multimedia); cs.NE (Neural and Evolutionary Computing); cs.NI (Networking and Internet Architecture); cs.PF (Performance); cs.PL (Programming Languages); cs.RO (Robotics); cs.SD (Sound); cs.SE (Software Engineering); cs.SI (Social and Information Networks); cs.SY (Systems and Control).
Selected 31 of 388 available items for this weekly brief.