Unverified50% confidenceFactTime unknown
GitHub无障碍智能体测试中已审查3535个PR，问题解决率达到68%
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
AI日报：OpenClaw性能提升3.5倍，Copilot Memory开放，编程工具全面进化
bilibili硅基米德AI
Related Claims
Unverified2023年普林斯顿大学的SWE-bench基准测试显示，当时最先进的AI编程工具在真实GitHub Issue修复任务中的成功率仅约13%75% similarUnverifiedGitHub's universal accessibility agent has reviewed 3,535 PRs with a 68% issue resolution rate75% similarUnverifiedGitHub与纽约大学联合研究（2021年发表于IEEE S&P）发现在测试的89个安全敏感场景中，GitHub Copilot约40%的建议包含安全漏洞73% similarUnverifiedAnything Analyzer在GitHub上获得了2197个Star和464个Fork70% similarUnverified2024年多项针对SWE-bench的评估显示，最先进的AI代理在真实GitHub Issue的解决率上仅在20%~50%区间70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/57165API
curl https://kongchang.com/api/v1/knowledge/claims/57165MCP
get_claim(id=57165)