新闻
What We Learned by Reproducing 2,200 papers from ICML
Hugging Face · 发布于 · 阅读约3分钟
30秒读懂
- 发生了什么
- A hackathon reproduced 2,226 ICML 2026 papers using coding agents, finding 51% had verified claims and 23% had falsified or contested claims.
- 为何重要
- Matters for researchers evaluating conference paper reliability and for those using AI agents to audit scientific work at scale.
- 注意
- Reproducibility is adversarial, not binary. Independent teams reached opposite verdicts on same claims. Missing artifacts limited many reproductions.
收听本摘要
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。