Unverified50% confidenceBenchmarkExact time
Meter analyzed AI pull requests that passed tests and found that approximately half would not be merged by project maintainers.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/6/2026
First Seen
Valid until: 1/4/2027
Sources
Related Entities
Related Claims
UnverifiedIn the experimental design, the AI used an 8:2 split to reserve 20% as an independent test set, and applied 5-fold cross-validation on the 80% training set to avoid data leakage.65% similarUnverifiedMost AI programming assistants currently only provide aggregated percentages and lack per-request consumption records.64% similarUnverifiedA MIT study published this year found that real project maintainers reject approximately half of AI-generated fixes when reviewing them, even when those fixes pass all tests.63% similarUnverified在某些测试条件下，主流AI编程工具推荐不存在包名的概率可达5%~20%62% similarUnverifiedAI生成假设的边际成本趋近于零，而实验验证成本受湿实验室物理约束无法同步降低，形成验证瓶颈60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/981978API
curl https://kongchang.com/api/v1/knowledge/claims/981978MCP
get_claim(id=981978)