Event ·Official first-hand

o4-mini评估：解决大学物理问题准确率约90%，图文问题仅79%

研究团队评估了OpenAI的o4-mini模型在Halliday和Resnick的《物理学基础》习题上的表现。整体准确率约90%，但文本问题准确率高达96%，而需要图文协调的问题仅79%。问题难度从低到高时，准确率显著下降。结果表明，当前最强推理模型能解答大部分标准物理题，但表现受模态和难度制约。

Current read

OpenAI的o4-mini做大学物理题整体准，但看图题就露馅了，准确率从96%掉到79%。

11 sources59° AI heatLast updated 2026/7/15 19:08:56

Evidence

11 sources

Primary sourcearXiv: OpenAI

2026/7/15 19:08:56View source

Related source 1AWS Machine Learning Blog

2026/7/13 21:01:20View source

Related source 2Greg Brockman

2026/7/14 03:56:16View source

Related source 3Sam Altman

2026/7/14 14:26:24View source

Related source 4IT之家

2026/7/15 03:00:22View source

Related source 5Decoder

2026/7/15 17:35:12View source

Related source 6shao__meng

2026/7/16 02:24:07View source

Related source 7宝玉

2026/7/13 21:33:40View source

Related source 8AI Will

2026/7/14 09:11:33View source

Related source 9@OpenAIDevs

2026/7/14 18:48:15View source

Related source 10OpenAI

2026/7/15 17:34:30View 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.

Open the story