Unverified70% confidenceFactTime unknown
目前顶尖AI Agent（如Devin、SWE-agent）在SWE-bench上的解决率约为40-55%
1
Sources
70%
Confidence
Medium-term (~90 days)
Relevance
6/1/2026
First Seen
Valid until: 8/30/2026
Sources
Crafta-Bench：Cursor后台Agent基准测试工具深度解析
githubnwinter
Related Entities
Related Claims
Unverified部分AI Agent系统在SWE-bench上的解题率已突破50%80% similarUnverified在AI Agent四大核心组件中，规划能力的成熟度最低（★★☆☆☆），大模型成熟度最高（★★★★★）72% similarUnverified当前AI Agent在SWE-bench测试中能解决80%的问题70% similarUnverified根据Lunchen开发团队对全球超过1300名专业人士的调查，57.3%的项目已经部署了AI Agent67% similarUnverified最强AI Agent在OSWorld 2.0上的任务完成率仅约20%67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/31097API
curl https://kongchang.com/api/v1/knowledge/claims/31097MCP
get_claim(id=31097)