Unverified90% confidenceFactTime unknown

SWE-bench（Software Engineering Benchmark）是目前衡量智能体编码能力的核心基准，从真实的GitHub Issue出发要求模型自主定位代码库中的问题并提交修复补丁

1

Sources

90%

Confidence

Long-term

Relevance

6/1/2026

First Seen

Sources

Claude 4.5 Sonnet实测：一条指令构建完整AI视觉应用

bilibiliUltralytics

Related Entities

SWE-benchconcept GitHubcompany

Related Claims

UnverifiedSWE-bench要求模型解决GitHub上真实的软件缺陷，是评估Agent级编程能力的权威基准79% similar VerifiedSWE-bench直接从GitHub真实开源项目中提取issue和对应的pull request，要求AI系统理解问题描述、定位代码缺陷并生成正确的修复补丁77% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/24179

API

curl https://kongchang.com/api/v1/knowledge/claims/24179

MCP

get_claim(id=24179)