Event ·Multi-source

CritPt v2 登陆 Artificial Analysis:研究级物理基准全面修订

Artificial Analysis 上线 CritPt v2,对 70 道研究级物理挑战全部修订,修正了表述不清、错误和笔误,并改进了评分器稳定性。模型现在以智能体方式作答,可在离线沙盒中编写并运行代码。榜单上 GPT-6 Astra (max) 以 83% 居首,GPT-6.1 Sol (max) 得 82%,Claude Opus 5.5 得 78%。GPT-5 (high) 在 CritPt v2 上仅得 29%,与 GPT-6 Astra 相差 54 个百分点。Intelligence Index 将于 10 月底更新并纳入 CritPt v2。

Current read

研究级物理基准 CritPt 出 v2 了,题目全改过,模型还能跑代码作答。看看 GPT-6 和 GPT-5 差了 54 分,前沿进展挺直观的。

11 sources68° AI heatLast updated 2026/10/22 17:55:37

Evidence

11 sources
Primary sourceArtificial Analysis
View source
Related source 1arXiv: DeepSeek
View source
Related source 2TestingCatalog
View source
Related source 3swyx (AI Engineer)
View source
Related source 4rohanpaul_ai
View source
Related source 8Teortaxes
View source
Related source 10Bindu Reddy
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.