Unverified50% confidenceFactExact time
The research team tested 22 open-source LLMs from six model families
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe team added 21 test cases for the config file writing mechanism.72% similarUnverifiedOpenAI released the GeneBench Pro research benchmark, comprising 129 evaluation tasks spanning 10 major domains and 21 sub-directions, with a top score of 31.5% among frontier models.64% similarUnverifiedIn actual testing, one Deep Research call invoked 104 agents for targeted research, completing fact-checking based on 22 sources and 59 claims.63% similarUnverified团队在项目期间经历四五次模型迭代，每次都发现模型能力越强从验证循环中获得的收益越大，对验证循环的需求是更持久的底层规律62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/893420API
curl https://kongchang.com/api/v1/knowledge/claims/893420MCP
get_claim(id=893420)