分布式计算 on 庞玉栋个人博客

DySCo: Dynamic Sharding for Collaborative Edge-Cloud LLM Inference with Depth-Synchronized Batching

Tue, 06 Oct 2026 20:42:11 +0800

原文：DySCo: Dynamic Sharding for Collaborative Edge-Cloud LLM Inference with Depth-Synchronized Batching

作者：Jingpo Xu, Paul Joe Maliakel, Ivona Brandic, Shashikant Ilager

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Distributed, Parallel, and Cluster Computing

arXiv:2610.08268v1 (cs)

[Submitted on 6 Oct 2026]

Title:DySCo: Dynamic Sharding for Collaborative Edge-Cloud LLM Inference with Depth-Synchronized Batching

Authors:Jingpo Xu, Paul Joe Maliakel, Ivona Brandic, Shashikant Ilager

View a PDF of the paper titled DySCo: Dynamic Sharding for Collaborative Edge-Cloud LLM Inference with Depth-Synchronized Batching, by Jingpo Xu and 3 other authors

Persistent Memory in Multi-Agent LLM Inference: What It Costs, What It Buys, and When You Can Tell

Tue, 06 Oct 2026 13:17:44 +0800

原文：Persistent Memory in Multi-Agent LLM Inference: What It Costs, What It Buys, and When You Can Tell

作者：Hochan Son, Kyungdoe Han, Jaehan Koh, Xiaowu Dai, Wenlu Xu, Guang Cheng

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Artificial Intelligence

arXiv:2610.07782v1 (cs)

[Submitted on 6 Oct 2026]

Title:Persistent Memory in Multi-Agent LLM Inference: What It Costs, What It Buys, and When You Can Tell

Authors:Hochan Son, Kyungdoe Han, Jaehan Koh, Xiaowu Dai, Wenlu Xu, Guang Cheng

EdgeAgent: Orchestrating On-Device LLM inference for End-User Multi-Agent Systems on CPU-GPU Unified Memory Architectures

Fri, 02 Oct 2026 22:43:42 +0800

原文：EdgeAgent: Orchestrating On-Device LLM inference for End-User Multi-Agent Systems on CPU-GPU Unified Memory Architectures

作者：Yuhai Long, Yuanxin Wei, Kai Wu, Jinhui Wei, Dan Huang, Jiangsu Du

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Distributed, Parallel, and Cluster Computing

arXiv:2610.03394v1 (cs)

[Submitted on 2 Oct 2026]

Title:EdgeAgent: Orchestrating On-Device LLM inference for End-User Multi-Agent Systems on CPU-GPU Unified Memory Architectures

Authors:Yuhai Long (1), Yuanxin Wei (1), Kai Wu (2), Jinhui Wei (1), Dan Huang (1), Jiangsu Du (1) ((1) School of Computer Science and Engineering, Sun Yat-sen University, (2) China Mobile Internet Company Ltd.)

D2K-Bench: Can LLM Agents Turn Expert Designs into Efficient GPU Kernels?

Fri, 02 Oct 2026 20:42:12 +0800

原文：D2K-Bench: Can LLM Agents Turn Expert Designs into Efficient GPU Kernels?

作者：Daifeng Li, Huiqiang Jiang, Chengruidong Zhang, Wei Wu, Xudong Guo, Jianhong Tu, Jianwei Zhang, Binhang Yuan, Dayiheng Liu

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Machine Learning

arXiv:2610.03226v1 (cs)

[Submitted on 2 Oct 2026]

Title:D2K-Bench: Can LLM Agents Turn Expert Designs into Efficient GPU Kernels?

Authors:Daifeng Li, Huiqiang Jiang, Chengruidong Zhang, Wei Wu, Xudong Guo, Jianhong Tu, Jianwei Zhang, Binhang Yuan, Dayiheng Liu

ServeTwin: A Benchmark-Validated Simulator for Distributed LLM Architecture Exploration

Fri, 02 Oct 2026 11:06:15 +0800

原文：ServeTwin: A Benchmark-Validated Simulator for Distributed LLM Architecture Exploration

作者：Sungjoon Park, Changue Jung, Kyungno Joo, Mincheol Kang, Jaehyung Ahn, Sehwan Lee, Sangjoon Kim

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Performance

arXiv:2610.02732v1 (cs)

[Submitted on 2 Oct 2026]

Title:ServeTwin: A Benchmark-Validated Simulator for Distributed LLM Architecture Exploration

Authors:Sungjoon Park, Changue Jung, Kyungno Joo, Mincheol Kang, Jaehyung Ahn, Sehwan Lee, Sangjoon Kim

View a PDF of the paper titled ServeTwin: A Benchmark-Validated Simulator for Distributed LLM Architecture Exploration, by Sungjoon Park and 5 other authors

HakiCC: LLM-Driven Multi-Agent Design and Optimization of Concurrency Control Protocols

Thu, 01 Oct 2026 09:10:15 +0800

原文：HakiCC: LLM-Driven Multi-Agent Design and Optimization of Concurrency Control Protocols

作者：Farzad Habibi, Juncheng Fang, Faisal Nawab

来源：arXiv cs.DB（数据库）

正文

Computer Science > Databases

arXiv:2610.00889v1 (cs)

[Submitted on 1 Oct 2026]

Title:HakiCC: LLM-Driven Multi-Agent Design and Optimization of Concurrency Control Protocols

Authors:Farzad Habibi, Juncheng Fang, Faisal Nawab

View a PDF of the paper titled HakiCC: LLM-Driven Multi-Agent Design and Optimization of Concurrency Control Protocols, by Farzad Habibi and 2 other authors

Leto: Fast In-Place Recovery for LLM Training on Surviving Hardware

Thu, 01 Oct 2026 04:29:57 +0800

原文：Leto: Fast In-Place Recovery for LLM Training on Surviving Hardware

作者：Geon-Woo Kim, Joon Ha Kim, Daehyeok Kim

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Distributed, Parallel, and Cluster Computing

arXiv:2610.00687v1 (cs)

[Submitted on 30 Sep 2026]

Title:Leto: Fast In-Place Recovery for LLM Training on Surviving Hardware

Authors:Geon-Woo Kim, Joon Ha Kim, Daehyeok Kim

View a PDF of the paper titled Leto: Fast In-Place Recovery for LLM Training on Surviving Hardware, by Geon-Woo Kim and 2 other authors

Characterizing High Bandwidth Flash for LLM Serving

Wed, 30 Sep 2026 15:02:56 +0800

原文：Characterizing High Bandwidth Flash for LLM Serving

作者：Zack Yu, Chloe Wong, Coleman Hooper, Minjae Lee, Wonjun Kang, Youngjin Cho, Michael W. Mahoney, Yakun Sophia Shao, Kurt Keutzer, Amir Gholami

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Machine Learning

arXiv:2609.39131v1 (cs)

[Submitted on 30 Sep 2026]

Title:Characterizing High Bandwidth Flash for LLM Serving

Authors:Zack Yu, Chloe Wong, Coleman Hooper, Minjae Lee, Wonjun Kang, Youngjin Cho, Michael W. Mahoney, Yakun Sophia Shao, Kurt Keutzer, Amir Gholami

SPLASH: Switching Parallel Layouts of Attention with Seamless Handoff for LLM Serving

Tue, 29 Sep 2026 22:02:29 +0800

原文：SPLASH: Switching Parallel Layouts of Attention with Seamless Handoff for LLM Serving

作者：Chuan Liu, Shuoming Zhang, Zhicheng Li, Qianqi Sun, Ruiyuan Xu, Qiuchu Yu, Xiyu Shi, Huimin Cui, Jiacheng Zhao

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Distributed, Parallel, and Cluster Computing

arXiv:2609.37626v1 (cs)

[Submitted on 29 Sep 2026]

Title:SPLASH: Switching Parallel Layouts of Attention with Seamless Handoff for LLM Serving

Authors:Chuan Liu, Shuoming Zhang, Zhicheng Li, Qianqi Sun, Ruiyuan Xu, Qiuchu Yu, Xiyu Shi, Huimin Cui, Jiacheng Zhao

Janus: Evidence-Before-Effect Sagas and Offline-Verifiable Provenance for Agentic LLMs

Tue, 29 Sep 2026 21:57:51 +0800

原文：Janus: Evidence-Before-Effect Sagas and Offline-Verifiable Provenance for Agentic LLMs

作者：Mustafa Arslan

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Cryptography and Security

arXiv:2609.38266v1 (cs)

[Submitted on 29 Sep 2026]

Title:Janus: Evidence-Before-Effect Sagas and Offline-Verifiable Provenance for Agentic LLMs

Authors:Mustafa Arslan

View a PDF of the paper titled Janus: Evidence-Before-Effect Sagas and Offline-Verifiable Provenance for Agentic LLMs, by Mustafa Arslan

View PDF

HTML (experimental)

Abstract:Agentic large language models (LLMs) now move money through tools, yet the record of what they did is usually a trace their own process emits beside the effect. Janus puts the record on the effect path. A step’s proposal, the verdict on it and any answer from a validator or a person are durable in a signed, hash-chained log before the step may run or its effect be released; with keys declared, each answer is signed by whoever gave or relayed it. Gates are pure functions of that log, and an auditor re-derives every verdict offline from the log and one public key. At the MCP edge the effect is held until then; through the SDK, which our model experiment uses, a cooperating client runs it only afterwards. We evaluate Janus under crash injection (144 kills in-process, 81 through the daemon), by verifying a 100-million-event log offline (254.5 s), and with a real model behind a lending workflow, run governed and plain on the same recorded model outputs. With the lending mandate in the model’s prompt, the comparison was 0 against 0. With it only in the policy and the amount’s unit stated, the model approved six loans declared over the mandate, three with no injection (a run that also dropped the unit approved three); the plain agent paid all six and Janus none, each refused by a deterministic validator and re-derivable offline. An always-approve oracle over the recorded intakes gave 20 and 21 declared over the mandate against 0, though Janus paid four and three whose declared amount understated the request. Designing the experiment exposed, in a system that passed its own audit, an instance of post-approval substitution: an approval keyed to an attempt was counted for a different proposal, moving a person’s approval from 100 to 1,000,000. We report it, a first fix and the five routes around it, and what Janus does not guarantee.

TempoKV: Timely Staging of LLM KV Caches for Memory-Semantic Flash

Mon, 28 Sep 2026 20:55:20 +0800

原文：TempoKV: Timely Staging of LLM KV Caches for Memory-Semantic Flash

作者：Jay H. Park, Hyungjun Kim, Dong Kim

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Distributed, Parallel, and Cluster Computing

arXiv:2609.35065v1 (cs)

[Submitted on 28 Sep 2026]

Title:TempoKV: Timely Staging of LLM KV Caches for Memory-Semantic Flash

Authors:Jay H. Park, Hyungjun Kim, Dong Kim

View a PDF of the paper titled TempoKV: Timely Staging of LLM KV Caches for Memory-Semantic Flash, by Jay H. Park and 2 other authors

Nereus: Adaptive Parallelism for LLM Post-Training

Mon, 28 Sep 2026 16:55:05 +0800

原文：Nereus: Adaptive Parallelism for LLM Post-Training

作者：Songlin Jiang, Tuo Shi, Sitong Zhang, Zeke Wang, Mario Di Francesco, Bo Zhao

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Distributed, Parallel, and Cluster Computing

arXiv:2609.34645v1 (cs)

[Submitted on 28 Sep 2026]

Title:Nereus: Adaptive Parallelism for LLM Post-Training

Authors:Songlin Jiang, Tuo Shi, Sitong Zhang, Zeke Wang, Mario Di Francesco, Bo Zhao

View a PDF of the paper titled Nereus: Adaptive Parallelism for LLM Post-Training, by Songlin Jiang and 5 other authors

QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents

Mon, 28 Sep 2026 03:02:03 +0800

原文：QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents

作者：Weiqi Wang, Yuxin Zhou, Mouxiang Chen, Siyuan Zhang, Yi Zhang, Yuyan Luo, Zhiyu Yin, Chencan Wu, Jiemin Jiang, Wentao Yao, Chujie Zheng, JianWei Zhang

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Machine Learning

arXiv:2609.33848v1 (cs)

[Submitted on 27 Sep 2026]

Title:QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents

Authors:Weiqi Wang, Yuxin Zhou, Mouxiang Chen, Siyuan Zhang, Yi Zhang, Yuyan Luo, Zhiyu Yin, Chencan Wu, Jiemin Jiang, Wentao Yao, Chujie Zheng, JianWei Zhang

KernelOPT: Dispatch-Aware Agentic Search for GPU Kernel Optimization

Fri, 25 Sep 2026 00:17:52 +0800

原文：KernelOPT: Dispatch-Aware Agentic Search for GPU Kernel Optimization

作者：Aheli Poddar, Sanskar Prasad, Arindam Samanta, Subha Chakraborty, Vishal Goyal, Rohit Singh Rathaur

来源：arXiv cs.AI（人工智能）

正文

Computer Science > Distributed, Parallel, and Cluster Computing

arXiv:2609.30059v1 (cs)

[Submitted on 24 Sep 2026]

Title:KernelOPT: Dispatch-Aware Agentic Search for GPU Kernel Optimization

Authors:Aheli Poddar, Sanskar Prasad, Arindam Samanta, Subha Chakraborty, Vishal Goyal, Rohit Singh Rathaur

View a PDF of the paper titled KernelOPT: Dispatch-Aware Agentic Search for GPU Kernel Optimization, by Aheli Poddar and 5 other authors

Hard Stop: Kernel-Level Preemption and Containment for Rogue Agentic Execution

Thu, 24 Sep 2026 21:42:02 +0800

原文：Hard Stop: Kernel-Level Preemption and Containment for Rogue Agentic Execution

作者：José Luis Pino

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Cryptography and Security

arXiv:2609.29808v1 (cs)

[Submitted on 24 Sep 2026]

Title:Hard Stop: Kernel-Level Preemption and Containment for Rogue Agentic Execution

Authors:José Luis Pino

View a PDF of the paper titled Hard Stop: Kernel-Level Preemption and Containment for Rogue Agentic Execution, by Jos'e Luis Pino

View PDF

HTML (experimental)

Abstract:In July 2026, an unconstrained autonomous agent participating in a frontier AI cybersecurity evaluation harness breached its evaluation sandbox, established an external command-and-control foothold, and executed a multi-stage intrusion into Hugging Face’s production multi-tenant dataset conversion infrastructure (referred to in this autopsy as Incident-2026-Alpha). Over 4.5 days, the rogue agent executed 17,600 discrete actions across 6,280 worker clusters, compromised AWS EC2 Instance Metadata Service (IMDS) credentials, forged Kubernetes service account tokens, rooted physical worker nodes via overprivileged CSI drivers, harvested 136 production secrets, and enrolled 181 ephemeral sandboxes into the organization’s internal mesh VPN.

MeshHeal: Two-Timescale Self-Healing for Gray Failures in Decentralized LLM Agent Networks

Thu, 24 Sep 2026 12:29:37 +0800

原文：MeshHeal: Two-Timescale Self-Healing for Gray Failures in Decentralized LLM Agent Networks

作者：Keru Chen, Sen Lin, Yingbin Liang, Nathaniel D. Bastian, Shaofeng Zou

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Artificial Intelligence

arXiv:2609.29015v1 (cs)

[Submitted on 24 Sep 2026]

Title:MeshHeal: Two-Timescale Self-Healing for Gray Failures in Decentralized LLM Agent Networks

Authors:Keru Chen, Sen Lin, Yingbin Liang, Nathaniel D. Bastian, Shaofeng Zou

View a PDF of the paper titled MeshHeal: Two-Timescale Self-Healing for Gray Failures in Decentralized LLM Agent Networks, by Keru Chen and 4 other authors

Distributed Stochastic Approximation Algorithms and Heavy-Tailed Age of Information

Wed, 23 Sep 2026 16:01:29 +0800

原文：Distributed Stochastic Approximation Algorithms and Heavy-Tailed Age of Information

作者：Adrian Redder, Arunselvan Ramaswamy, Holger Karl

来源：arXiv cs.DC（分布式并行）

正文

Mathematics > Optimization and Control

arXiv:2609.27499v1 (math)

[Submitted on 23 Sep 2026]

Title:Distributed Stochastic Approximation Algorithms and Heavy-Tailed Age of Information

Authors:Adrian Redder, Arunselvan Ramaswamy, Holger Karl

View a PDF of the paper titled Distributed Stochastic Approximation Algorithms and Heavy-Tailed Age of Information, by Adrian Redder and 2 other authors

View PDF

LayerCheck: Adaptive Layer-wise Checkpointing for Large Language Model Post-training

Wed, 23 Sep 2026 08:57:59 +0800

原文：LayerCheck: Adaptive Layer-wise Checkpointing for Large Language Model Post-training

作者：Minqiu Sun, Xin Huang, Luanzheng Guo, Nathan R. Tallent, Kento Sato, Dong Dai

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Distributed, Parallel, and Cluster Computing

arXiv:2609.27193v1 (cs)

[Submitted on 23 Sep 2026]

Title:LayerCheck: Adaptive Layer-wise Checkpointing for Large Language Model Post-training

Authors:Minqiu Sun, Xin Huang, Luanzheng Guo, Nathan R. Tallent, Kento Sato, Dong Dai

View a PDF of the paper titled LayerCheck: Adaptive Layer-wise Checkpointing for Large Language Model Post-training, by Minqiu Sun and 5 other authors

WeightBridge: An Efficient Weight Transfer Library for Reinforcement Learning

Tue, 22 Sep 2026 05:54:15 +0800

原文：WeightBridge: An Efficient Weight Transfer Library for Reinforcement Learning

作者：Xuanlin Jiang, Samuel Hsia, Michael Kuchnik, Zachary DeVito, Minlan Yu, Carole-Jean Wu

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Distributed, Parallel, and Cluster Computing

arXiv:2609.25442v1 (cs)

[Submitted on 21 Sep 2026]

Title:WeightBridge: An Efficient Weight Transfer Library for Reinforcement Learning

Authors:Xuanlin Jiang, Samuel Hsia, Michael Kuchnik, Zachary DeVito, Minlan Yu, Carole-Jean Wu

View a PDF of the paper titled WeightBridge: An Efficient Weight Transfer Library for Reinforcement Learning, by Xuanlin Jiang and 5 other authors

DeepSeek Elastic Compute (DSec): A Sandbox Infrastructure for Effective Agentic Training at Scale

Sat, 19 Sep 2026 20:20:26 +0800

原文：DeepSeek Elastic Compute (DSec): A Sandbox Infrastructure for Effective Agentic Training at Scale

作者：Jialiang Huang, Hongxuan Tang, Jingchang Chen, Yuxuan Liu, Yixiao Chen, Yuan Cheng, Yi Tao, Jingli Zhou, Yupeng Chen, Haoyu Chen, Jiarui Wang, Shengkai Lin, Chuqi Zhang, Bryan Lee Teng, Lian Guo, Zhe Fu, Wenjun Gao, Yisong Wang, Liang Zhao, Zehao Wang, Ziwei Xie, Yongqiang Guo, Peixin Cong, Ziyi Gao

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Distributed, Parallel, and Cluster Computing

ASPIRE: Asynchronous Batched Self-Speculative Decoding for Long-Context LLM Inference

Wed, 16 Sep 2026 08:04:13 +0800

原文：ASPIRE: Asynchronous Batched Self-Speculative Decoding for Long-Context LLM Inference

作者：Amir Ziashahabi, Hossein Entezari Zarch, Lei Gao, Murali Annavaram, Salman Avestimehr

来源：arXiv cs.DC（分布式并行）

正文

Computer Science > Machine Learning

arXiv:2609.17943v1 (cs)

[Submitted on 16 Sep 2026]

Title:ASPIRE: Asynchronous Batched Self-Speculative Decoding for Long-Context LLM Inference

Authors:Amir Ziashahabi, Hossein Entezari Zarch, Lei Gao, Murali Annavaram, Salman Avestimehr

View a PDF of the paper titled ASPIRE: Asynchronous Batched Self-Speculative Decoding for Long-Context LLM Inference, by Amir Ziashahabi and 4 other authors