Research Papers
Research Papers news — Friday 16 October 2026
29 stories from 2 sources, every one in English. Each links to its publisher.
-
Local Epochs as Optimal Stopping: Theory and Evidence in Federated Learning
arXiv cs.DC (Distributed)
arXiv:2610.21832v1 Announce Type: cross Abstract: How long should each client train before communicating? Federated learning usually answers with a fixed, hand-tuned number of local epochs, yet an extra epoch helps only while the loss it…
-
When the KV Cache Is Slower Than the GPU: Scaling Remote KV Tiers for Context-Parallel Prefill
arXiv cs.DC (Distributed)
arXiv:2610.21500v1 Announce Type: new Abstract: Reusing the attention KV cache across requests is a standard way to cut prefill cost in long-context LLM serving, and production systems increasingly back the per-engine cache with a remote…
-
Helix: A Unified Programmable Accelerator for Diverse Krylov Subspace Solvers
arXiv cs.DC (Distributed)
arXiv:2610.21096v1 Announce Type: cross Abstract: Domain-specific accelerators deliver high performance and energy efficiency through specialization but often lack support for extensible algorithm families. Existing FPGA accelerators for…
-
Efficient Heuristics and Machine Learning Approach for Fault Characterization in Distributed Self-Stabilizing Programs
arXiv cs.DC (Distributed)
arXiv:2610.10386v2 Announce Type: replace Abstract: Modern large-scale systems rely on distributed protocols to maximize efficiency while preserving the correctness guarantees of single-process execution. However, designing such protocols…
-
Efficient KV Cache Migration for Geo-Distributed LLM Inference in Collaborative Edge Computing
arXiv cs.DC (Distributed)
arXiv:2610.22172v1 Announce Type: new Abstract: Large Language Models (LLMs) have shown exceptional capabilities but rely heavily on centralized cloud GPU clusters, leading to high bandwidth costs, latency, and privacy issues. Edge…
-
Harnessing MPI mutations for AI error detection
arXiv cs.DC (Distributed)
arXiv:2610.20616v1 Announce Type: cross Abstract: MPI errors are challenging to identify despite the significant number of expert verification tools. Dynamic tools (i.e., requiring profiling) are computationally expensive and accurate in…
-
Architecting for Hardware Diversity: Meta's End-to-End Co-Design for High-Performance AI Training on AMD GPUs
arXiv cs.DC (Distributed)
arXiv:2610.22230v1 Announce Type: new Abstract: Meta deploys a heterogeneous fleet comprising NVIDIA GPUs, AMD GPUs, and proprietary MTIA accelerators to reduce costs and mitigate vendor lock-in. However, the extensive history of…
-
Fast TetraBFT: Optimizing Latency Where It Matters
arXiv cs.DC (Distributed)
arXiv:2606.03754v2 Announce Type: replace Abstract: Unauthenticated Byzantine consensus protocols achieve optimal failure resilience while relying only on authenticated point-to-point channels, not authenticated messages. They are an…
-
A Model-Centric DevOps Architecture for DEVS-Based Digital Twin Simulation Services
arXiv cs.DC (Distributed)
arXiv:2609.11122v2 Announce Type: replace-cross Abstract: Digital twin simulation models evolve and are redeployed like software. Engines based on the Discrete Event System Specification (DEVS) have a sound formal basis but offer little…
-
NosRacer: Dynamic Detection of Race Conditions in On-Device Network Operating Systems
arXiv cs.DC (Distributed)
arXiv:2610.11875v2 Announce Type: replace Abstract: Commercial on-device network operating systems (NOSes) run the complex control planes of production routers and switches, where configuration update tasks are executed by multiple loosely…
-
Coexponential Exploration in the Clef Programming Language
arXiv cs.DC (Distributed)
arXiv:2610.12911v2 Announce Type: replace-cross Abstract: Clef was created as a concurrent programming language whose compiler tracks state ownership, waiting computations and request dependencies. We develop a coexponential account of…
-
ZenFlow: Enabling Stall-Free Offloading for LLM Training
arXiv cs.DC (Distributed)
arXiv:2505.12242v4 Announce Type: replace Abstract: Fine-tuning large language models (LLMs) often exceeds GPU memory limits, prompting systems to offload model states to CPU memory. However, existing offloaded training frameworks like…
-
TensorDex: A Compact, Tensor-Centric Storage System for Modern AI Models
arXiv cs.DC (Distributed)
arXiv:2604.17104v3 Announce Type: replace Abstract: Modern model hubs store hundreds of petabytes of large language models (LLMs), with fine-tuned variants dominating the storage footprint. These variants contain substantial cross-model…
-
Asynchronous Indulgent Gathering of Opaque Robots
arXiv cs.DC (Distributed)
arXiv:2509.10711v2 Announce Type: replace Abstract: We consider the benchmarking problem of gathering in an $(N,f)$-fault system consisting of $N$ robots, of which at most $f$ may fail at any execution. The {\em traditional} version asks…
-
TeraAgent: A Distributed Agent-Based Simulation Engine for Simulating Half a Trillion Agents
arXiv cs.DC (Distributed)
arXiv:2509.24063v2 Announce Type: replace Abstract: Agent-based simulation is a method for studying highly complex systems in disciplines as diverse as biology, medicine, sociology, economics, finance, and more. In this method, an agent…
-
Cyclotron: A Recurrence Language for Expressing Interprocessor Communication
arXiv cs.DC (Distributed)
arXiv:2610.20981v1 Announce Type: new Abstract: We present Cyclotron, a compiler that expresses streaming dataflow algorithms using recurrence equations, which are then compiled to distributed topologies of interlinked processors. Our…
-
ReFlow: Replay-Driven Decoupling for Efficient RL Post-Training
arXiv cs.DC (Distributed)
arXiv:2610.21640v1 Announce Type: cross Abstract: Reinforcement learning (RL) has been a critical paradigm for improving large language model (LLM) reasoning and agentic capabilities, but its efficiency remains a major bottleneck. Existing…
-
AdaBool: Sparse Boolean Matrix Processing with Adaptive Data Representation
arXiv cs.DC (Distributed)
arXiv:2610.21264v1 Announce Type: cross Abstract: Many data processing tasks revolve around determining whether logical relationships exist between two sets of entities, a pattern that can be naturally modeled as Sparse Boolean Matrix…
-
ETH: Data-Balanced Heterogeneous Pipeline Parallelism for Training Large Models
arXiv cs.DC (Distributed)
arXiv:2610.22035v1 Announce Type: new Abstract: Current heterogeneity-aware training systems for large models balance workloads by assigning unequal numbers of layers across pipeline stages, a strategy aimed at equalizing execution time…
-
Resource-aware Sub-network Allocation for Heterogeneous Decentralized Learning
arXiv cs.DC (Distributed)
arXiv:2610.22015v1 Announce Type: new Abstract: System heterogeneity in computation, communication, and memory can substantially limit synchronous decentralized learning, where each training round is constrained by resource-limited nodes…
-
HyCo: Hybrid Drafting for Cost-Efficient Mobile-Cloud Speculative Decoding
arXiv cs.DC (Distributed)
arXiv:2610.21436v1 Announce Type: cross Abstract: Device-cloud speculative decoding reduces server computation, but growing a deep draft tree with a small language model (SLM) requires repeated passes on a mobile neural processing unit…
-
Probabilistic Optical-Layer Surrogate for Risk-Aware Control-Plane Decision Making
arXiv cs.DC (Distributed)
arXiv:2610.21069v1 Announce Type: new Abstract: In this paper, we propose a generative surrogate for the optical layer. Current control-plane decisions in optical transport networks are made using analytical impairment models…
-
Exact Makespan Bounds and the GPipe Homogeneity Bias for Heterogeneous LLM Inference Pipelines
arXiv cs.DC (Distributed)
arXiv:2610.21815v1 Announce Type: new Abstract: We give an exact critical-path characterization of the makespan of micro-batch pipelines serving LLMs on heterogeneous GPU clusters, quantifying how the classic homogeneous GPipe formula…
-
Volition Elicitation: Operational Semantics for People and Their Machines
arXiv cs.DC (Distributed)
arXiv:2607.14138v2 Announce Type: replace-cross Abstract: The most prevalent distributed systems today include people and their personal machines (smartphones), and computations in them are driven by people's volitions. Volition-guarded…
-
Coordination-Free Workload Regularization for Variable-Cost Distributed Training
arXiv cs.DC (Distributed)
arXiv:2610.21162v1 Announce Type: new Abstract: Synchronous data-parallel training usually assigns every rank the same number of records, even when record costs vary by orders of magnitude. For variable-length models, this converts input…
-
PrefixFlow: Training-Time KV Caching for Schedule-Level Prefix Reuse in LLM RL Training
arXiv cs.DC (Distributed)
arXiv:2606.01143v4 Announce Type: replace Abstract: In reinforcement learning (RL) post-training, one prompt fans out into $N$ trajectories, yet standard trainers replay the shared prefix's forward and backward pass $N$ times; in…
-
Benchmarking Post-Quantum Key Encapsulation in S3-Compatible Object Storage under Client Concurrency
arXiv cs.DC (Distributed)
arXiv:2610.21457v1 Announce Type: cross Abstract: Data encrypted today with classical public-key algorithms could be decrypted once large quantum computers exist. Cloud object stores keep data for years, so the client-side encryption that…
-
Wattlytics: A Web Platform for Co-Optimizing Performance, Energy, and TCO in HPC Clusters
arXiv cs.DC (Distributed)
arXiv:2604.08182v2 Announce Type: replace Abstract: The escalating computational demands and energy footprint of GPU-accelerated computing systems complicate informed design and operational decisions. We present the first release of…
-
Water treatment research opportunities at synchrotron and neutron user facilities
Nature Technology