ICML 2026 大规模复现挑战：AI 智能体验证 2200 篇论文
原标题：researchcommunityagents What We Learned by Reproducing 2,200 papers from ICML 5 August 13, 2026
AI 摘要
Hugging Face 社区在 2026 年 7 月 15 日至 8 月 2 日举办了 ICML 2026 开放复现挑战，参与者使用 AI 智能体复现了 2200 篇论文。结果显示，51% 的论文至少有一个声明被独立验证，23% 的论文有声明被证伪或质疑。挑战还确认了多起论文错误，包括证明缺陷、理论不匹配和评估偏差。该活动表明，AI 智能体可以大规模辅助科学验证，但也凸显了可复现性的复杂性。
正文节选
Reproduction: Towards Optimal Robustness in Learning-Augmented Paging Explore project logs, code, and traces in an interactive logbook In this post, we're sharing what we learned from running this hackathon, and what it suggests about the role humans will play when agents are doing the research experiments. Questions about how reproducible AI research really is are older than the current AI wave. But these questions are exacerbated by scale. ICML 2026 received 23,918 submissions and accepted 6,3