Unverified50% confidenceFactExact time
Metr（前身为ARC Evals）专注于前沿模型自主性与危险能力评估，曾在GPT-4和Claude 2发布前参与预发布测试
1
Sources
50%
Confidence
Long-term
Relevance
7/10/2026
First Seen
Sources
Related Claims
UnverifiedMETR evolved from the original ARC Evals team, which participated in third-party red-team evaluations of frontier models such as GPT-4 and Claude 2.73% similarUnverifiedGPT-5.2 topped the ARC-AGI benchmark.73% similarUnverifiedThe benchmark tested leading AI models including Claude, GPT series, and Gemini.72% similarUnverifiedModels such as GPT-4 can automatically generate structured test cases based on API documentation or requirement descriptions, covering boundary values and exception paths.71% similarUnverifiedOpenAI在GPT-4技术报告中已隐晦提及模型参与自身评估流程71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/464242API
curl https://kongchang.com/api/v1/knowledge/claims/464242MCP
get_claim(id=464242)