Event ·Official first-hand

PHYRD 基准发布:测试模型复现与扩展理论物理研究的能力

arXiv 论文介绍 PHYRD,一个针对理论物理推理的动态基准,包含 63 道 Reproduce 和 43 道 Discover 题。Reproduce 要求推导已发表结果,Discover 要求在修改理论条件后推导新结论,任务由 FormulaForest 系统构建并附带子问题级评分标准。五个模型在 Reproduce 上的整题准确率为 49.21% 到 68.25%,在闭卷 Discover 上为 18.60% 到 74.42%。使用工具后四个模型完成更多 Discover 题,但仍存在概念性错误和论证缺失。FormulaForest 还可用作合成数据,将 Qwen-35B-A3B 在 OpenAI 的 FrontierScience 基准上的准确率从 61% 提升到 72%。

Current read

物理版科研能力测试题,63 道复现加 43 道发现题,最强模型闭卷也只有七成多正确率,还有合成数据能白嫖涨 11 个点。

10 sources47° AI heatLast updated 2026/10/23 13:03:48

Evidence

10 sources
Primary sourcearXiv: OpenAI
View source
Related source 1The Information
View source
Related source 2Artificial Analysis
View source
Related source 3rohanpaul_ai
View source
Related source 4kimmonismus
View source
Related source 5techcrunch
View source
Related source 6VnExpress Số hóa
View source
Related source 7Bindu Reddy
View source
Related source 8karminski-牙医 (AI工具)
View source
Related source 9Nathan Lambert
View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.