Research Papers
Research Papers news — Monday 05 October 2026
33 stories from 2 sources, every one in English. Each links to its publisher.
-
Compound AI System Reliability: A Failure Taxonomy and Resilience Pattern Catalog from 150 Production Incidents
arXiv cs.DC (Distributed)
arXiv:2610.02503v1 Announce Type: cross Abstract: Deploying compound AI systems reliably and safely requires understanding failure modes that emerge at component boundaries, not within individual models. Cascading errors propagate across…
-
PaxosLease in Relativistic Inertial Frames
arXiv cs.DC (Distributed)
arXiv:2610.02879v1 Announce Type: new Abstract: A lease grants one holder exclusive authority over a resource for a fixed term, and its safety property guarantees that at most one holder has the lease at any time. Under special relativity…
-
Calibrating the Checksum: An Empirical False-Positive Bound for One-Sided ABFT on bfloat16 GEMM, and What It Fails to Catch
arXiv cs.DC (Distributed)
arXiv:2610.02240v1 Announce Type: cross Abstract: Algorithm-based fault tolerance detects corrupted matrix multiplications by comparing a checksum of the product against a checksum recomputed from the factors, flagging a fault when the…
-
Real-time Optimization of Simulation and Data Processing Pipelines for Experiments on Exascale Computing Platforms
arXiv cs.DC (Distributed)
arXiv:2610.02498v1 Announce Type: new Abstract: Current and next-generation large-scale experiments, especially in high-energy physics and cosmology, involve increasing volumes of data that require exascale computing power to analyze and…
-
Provisioning to Runtime Optimization of a 100 MW-Scale AI Cluster
arXiv cs.DC (Distributed)
arXiv:2605.24461v3 Announce Type: replace-cross Abstract: The electric power supply for AI data centers is now the most significant bottleneck in the race toward Artificial General Intelligence, surpassing even the constraint of AI…
-
PaNGEA: Parallel Node Generation and Exploration Algorithm on GPU
arXiv cs.DC (Distributed)
arXiv:2610.03090v1 Announce Type: new Abstract: Primal heuristics for finding high-quality feasible solutions are an important component in mixed-integer optimization (MIO) solvers. Recent advances in GPU-accelerated optimization…
-
Cross-Facility LLM Pre-training on HPC: Elastic Aggregation, Data Leasing, and Queue-Aware Placement
arXiv cs.DC (Distributed)
arXiv:2610.03457v1 Announce Type: new Abstract: Academic compute is fragmented: allocations are granted per facility, and facilities differ in accelerators and software stacks, schedule jobs independently, and share neither a network nor a…
-
Byzantine Causal Reliable Broadcast with Constant Metadata Overhead
arXiv cs.DC (Distributed)
arXiv:2609.37913v2 Announce Type: replace Abstract: Asynchronous Byzantine Reliable Broadcast (BRB) is a fundamental primitive that guarantees agreement and validity in distributed systems subject to Byzantine faults, but it lacks ordering…
-
ByteSplat: Efficient Distributed 3D Gaussian Splatting Training via Intra- and Inter-GPU communication reduction
arXiv cs.DC (Distributed)
arXiv:2610.02851v1 Announce Type: new Abstract: 3D Gaussian Splatting (3DGS) enables photorealistic scene reconstruction, but training large-scale scenes requires substantial memory and computation. Distributing training across multiple…
-
AFORE: Attention-FFN Disaggregation with Overlapped Reconfiguration of Experts
arXiv cs.DC (Distributed)
arXiv:2610.03203v1 Announce Type: new Abstract: Efficient serving of Mixture-of-Experts (MoE) models is challenging due to large expert parameters, input-dependent expert activation, and dynamic workloads. Expert parallelism distributes…
-
WakeKV: Reactive, Reversible KV Residency for Heads That Change Their Minds
arXiv cs.DC (Distributed)
arXiv:2610.02713v1 Announce Type: cross Abstract: Most KV-cache compression methods classify attention heads once, either offline or during prefill, and keep this classification fixed throughout generation. Across three models (1.5B-8B)…
-
A Path Integral Surrogate for Multi-Step Gradient Inversion in Federated Learning
arXiv cs.DC (Distributed)
arXiv:2610.03597v1 Announce Type: cross Abstract: Federated learning lets many clients train a shared model together without ever sending their private data to a central server. Each client shares only a model update, and this update…
-
D2K-Bench: Can LLM Agents Turn Expert Designs into Efficient GPU Kernels?
arXiv cs.DC (Distributed)
arXiv:2610.03226v1 Announce Type: cross Abstract: GPU kernels generated by large language model (LLM) agents can remain less efficient than expert implementations, but runtime alone does not reveal how the gap relates to design discovery…
-
VenusRL: A Fully Disaggregated Agentic RL System with Priority Scheduling and Scalable Interaction
arXiv cs.DC (Distributed)
arXiv:2610.03286v1 Announce Type: new Abstract: Agentic Reinforcement Learning (RL) trains LLM agents through multi-turn interactions with external tool environments. Its multi-turn nature exposes two system-level bottlenecks unaddressed…
-
Annotation-Driven Migration of CUDA Programs to Tenstorrent Blackhole
arXiv cs.DC (Distributed)
arXiv:2610.02658v1 Announce Type: new Abstract: Tenstorrent Blackhole combines distributed local memories, explicit inter-core communication, and Tensix cores decoupling data movement from computation. CUDA offers a substantial HPC…
-
ARGOS: Reinforcement Learning-Driven Multidimensional Elasticity for Service Orchestration in the Computing Continuum
arXiv cs.DC (Distributed)
arXiv:2609.37085v2 Announce Type: replace Abstract: Data-intensive services in the Computing Continuum must balance analytics quality, resource usage, and cost across heterogeneous nodes with limited and uneven capacity. This balance…
-
Coda: Exploiting Admission Flexibility for Coding-Agent Serving
arXiv cs.DC (Distributed)
arXiv:2610.03088v1 Announce Type: new Abstract: Coding agents powered by large language models (LLMs) repeatedly alternate between model inference and tool calls, creating long-lived sessions with reusable key-value (KV) states and…
-
EdgeAgent: Orchestrating On-Device LLM inference for End-User Multi-Agent Systems on CPU-GPU Unified Memory Architectures
arXiv cs.DC (Distributed)
arXiv:2610.03394v1 Announce Type: new Abstract: Emerging multi-agent LLMs demand privacy-preserving edge deployment, yet current inference systems struggle with these collaborative workflows. Specifically, the memory-bound decode phase…
-
ServeTwin: A Benchmark-Validated Simulator for Distributed LLM Architecture Exploration
arXiv cs.DC (Distributed)
arXiv:2610.02732v1 Announce Type: cross Abstract: Evaluating distributed LLM serving designs on physical clusters is costly. Yet existing simulators provide only subsets of the capabilities needed for realistic design exploration: stateful…
-
Learning to Rank Tensor Network Contraction Plans for GPU-Accelerated Quantum Circuit Simulation
arXiv cs.DC (Distributed)
arXiv:2608.05819v2 Announce Type: replace-cross Abstract: Classical simulation remains essential for developing and validating quantum algorithms, but its cost grows rapidly with circuit size. Tensor-network contraction can reduce this…
-
Closing the Prediction Gap: Completing Machine Shape So That Predicted Time, Power, Energy, and Mapping Match What Real Hardware Does
arXiv cs.DC (Distributed)
arXiv:2610.03197v1 Announce Type: new Abstract: A companion paper derives a weighted, communication free partition for heterogeneous, multi institution device ensembles from each device measured shape, rho_machine(d), validated on real…
-
RailWave: Adaptive Spatial and Temporal Scheduling for Expert-Parallel Communication
arXiv cs.DC (Distributed)
arXiv:2610.03415v1 Announce Type: new Abstract: Irregular All-to-All communication is a major bottleneck in expert-parallel Mixture-of-Experts (MoE) models. Even with fixed expert routing and placement, uneven utilization of parallel…
-
Feature Freshness Budgets for Real-Time ML Inference Under Stream Lag
arXiv cs.DC (Distributed)
arXiv:2610.02259v1 Announce Type: new Abstract: Online feature stores decouple feature computation from model serving, materializing features from upstream event streams into a low-latency store that inference reads at request time. This…
-
AI-Assisted GPU optimization of the Stochastic Variational Method for Few-Body Boson System
arXiv cs.DC (Distributed)
arXiv:2610.02745v1 Announce Type: cross Abstract: The stochastic variational method (SVM) is one of the most powerful methods to solve quantum few-body systems precisely in various fields, such as nuclear and atomic physics. To the best of…
-
Characterization-Guided GPU Fault Resilience in NVIDIA MPS
arXiv cs.DC (Distributed)
arXiv:2605.26461v2 Announce Type: replace Abstract: NVIDIA Multi-Process Service (MPS) enables fine-grained GPU sharing by allowing multiple processes to execute concurrently on the same GPU, making it an important mechanism for improving…
-
OpenMP Meta-Lowering: A Declarative Approach to Performance Portable Parallel Code Generation
arXiv cs.DC (Distributed)
arXiv:2610.03571v1 Announce Type: cross Abstract: The increasing diversity of parallel hardware challenges existing compilation flows. While OpenMP provides a portable abstraction for shared-memory parallelism, existing compilers tightly…
-
Distributed Variational Quantum Linear Solver
arXiv cs.DC (Distributed)
arXiv:2604.01426v2 Announce Type: replace-cross Abstract: This paper develops a distributed variational quantum algorithm for solving large-scale linear equations. For a linear system of the form $Ax=b$, the large square matrix $A$ is…
-
Lightweight and Resource-Efficient Perception for Robotic Guide Dogs
arXiv cs.DC (Distributed)
arXiv:2610.03187v1 Announce Type: new Abstract: Multi-camera streaming perception is increasingly deployed on heterogeneous edge platforms shared with co-resident workloads, yet accelerator placement is often evaluated using isolated…
-
Grassroots Bonds: Financing by the People, for the People
arXiv cs.DC (Distributed)
arXiv:2603.13671v4 Announce Type: replace Abstract: Grassroots currencies turn mutual trust into liquidity: a grassroots coin is a unit of its issuer's debt, backed by the issuer's goods and services, which the issuer must redeem, 1-for-1…
-
From Alert Floods to Precedence Forests: Zero-Prior-Knowledge Incident Triage with LOGOS
arXiv cs.DC (Distributed)
arXiv:2610.02297v1 Announce Type: new Abstract: Commercial observability platforms rely on domain artifacts like distributed traces, topology maps, and baseline metrics. However, when troubleshooting proprietary software, enterprise…
-
Beaver: Elastic GPU Sharing between ML and Latency-Critical vRAN Workloads
arXiv cs.DC (Distributed)
arXiv:2610.02522v1 Announce Type: new Abstract: Within the shared industry vision of AI-RAN, AI-and-RAN seeks to co-locate virtualized radio access network (vRAN) workloads and AI services on shared GPUs. This sharing is inherently…
-
Network-in-the-Loop at Scale: GPU-Batched 5G Simulation for Massively Parallel Robot Learning
arXiv cs.DC (Distributed)
arXiv:2610.02370v1 Announce Type: cross Abstract: Massively parallel GPU simulators train multi-robot policies in thousands of environments, and many fleets use private Fifth-Generation (5G) networks, where each robot's delay depends on…
-
How big tech is building a military–industrial complex in the age of AI
Nature Technology