今日速览
| 序号 | 标题 | 来源 | 日期 | 主题 | 推荐等级 |
|---|---|---|---|---|---|
| 1 | KREX: Concurrent Kernel Benchmarking on Shared GPUs via Region-Granular Exclusivity | arXiv | 2026-09-24 | AI-Agent | 高 |
| 2 | How does Adversarial Influence Scale in Multi-Agent Systems? | arXiv | 2026-09-24 | AI-Agent | 高 |
重点论文与技术动态
1. KREX: Concurrent Kernel Benchmarking on Shared GPUs via Region-Granular Exclusivity
- 来源:arXiv
- 日期:2026-09-24
- 作者/机构:Tianyu Feng, Haoxuan Yu, Tianyuan Wu, Lingyun Yang, Daocheng Ying, Yuxiao Wang
- 主题标签:
AI-Agent,arXiv - 推荐等级:高
- 分类:cs.DC, cs.OS, cs.PF
一句话结论
KREX 是共享 GPU 的并发内核基准测试运行时,以区域级独占保真、区域外并发,提升智能体测试吞吐。
核心内容
- 现有系统常整会话或整命令独占 GPU,需独占时间少,利用率低。
- KREX 标记计时敏感区域,仅区域内独占,区域外并发。
- 复用持久 GPU 上下文,避免重复的节点级串行创建。
方法与数据
- 区域内阻止新 GPU 提交、排空未完成工作,再冻结兄弟进程并隔离 CPU 核心。
- 在 NVIDIA/AMD GPU 上,较命令级独占基线最高 3.4 倍吞吐;p95 计时膨胀 0.30%、1.58%、3.90%,对应 >10 ms、>1 ms、>0.1 ms。
价值判断
- 值得关注:兼顾共享 GPU 吞吐与测量保真,适合智能体反复测试内核。
- 可复用点:区域标记、区域内独占、持久上下文复用、进程冻结与核心隔离。
- 局限/待核查:摘要未明确区域标记自动化、更短内核泛化性及节点规模。
摘要
LLM agents automate GPU kernel optimization by repeatedly composing candidates and measuring their duration on real GPUs. Existing systems preserve measurement fidelity by reserving a GPU for an entire agent session or benchmarking command. However, this results in poor utilization because only a small fraction of command execution requires exclusive GPU access. Sharing GPUs could recover this idle capacity, but introduces contention that compromises measurement fidelity and misdirects the agent's search. We present KREX, a runtime for concurrent kernel agent benchmarking with region-granular exclusivity. KREX lets agents mark critical regions involving timing-sensitive operations within a benchmarking command. The runtime then enforces exclusivity within marked regions and allows concurrent execution outside them, achieving high throughput while preserving measurement fidelity. To enforce in-region exclusivity, KREX blocks new competing GPU submissions and drains outstanding work before freezing sibling processes and isolating CPU cores, protecting both GPU execution and the host threads that drive measurements. To maximize off-region concurrency, KREX reuses GPU contexts in persistent context processes to avoid repeated, node-wide serialized context creation. We evaluate KREX on NVIDIA and AMD GPUs. Compared with command-granular exclusivity baselines, KREX delivers up to $3.4\times$ the benchmarking throughput with a negligible p95 timing inflation of $0.30\%$, $1.58\%$, and $3.90\%$ for kernels longer than 10 ms, 1 ms, and 0.1 ms, respectively.2. How does Adversarial Influence Scale in Multi-Agent Systems?
- 来源:arXiv
- 日期:2026-09-24
- 作者/机构:Addison J. Wu, Jasin Cekinmez, Michel Liao, Karthik Narasimhan, Thomas L. Griffiths
- 主题标签:
AI-Agent,arXiv - 推荐等级:高
- 分类:cs.AI, cs.CY
一句话结论
多智能体系统中,抗欺骗能力取决于欺骗者比例而非智能体总数;增加智能体数量不足以防御,因为对手可随群体扩展。
核心内容
- 研究群体规模增大、欺骗者更普遍时,多智能体讨论的易感性如何变化。
- 初始正确智能体最终改错的比例随欺骗者比例线性上升。
- 与人类从众实验相比,LLM 智能体在欺骗者仍为少数时也常改错;易感性还取决于交互模型,尤其诚实方模型;欺骗者私下协调反而可能降低效果。
方法与数据
- 摘要未明确实验设置、模型组合、任务与数据规模。
- 摘要未明确是否使用真实人类数据或仅 LLM 实验。
价值判断
- 值得关注:揭示多智能体“加人”不等于更安全,对 LLM 群体决策有警示。
- 可复用点:以“初始正确者改错率”衡量易感性,并关注欺骗者比例与模型组合。
- 局限/待核查:摘要未给出具体模型、任务、统计显著性与可复现数据。