Unverified50% confidenceBenchmarkExact time
SWE-bench is an authoritative benchmark for evaluating the capabilities of AI programming tools; it extracts tasks from real GitHub Issues and requires agents to autonomously locate code, fix bugs, and pass tests.
1
Sources
50%
Confidence
Long-term
Relevance
8/31/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedSWE-Agent is a research project released by Princeton University focused on enabling AI agents to autonomously solve real software engineering problems on GitHub.78% similarUnverifiedSWE-RPG是在SWE-bench基础上发展而来的新一代AI编程能力评估框架，引入了角色扮演机制模拟不同经验水平开发者的需求提交方式76% similarUnverifiedTerminal-Bench 属于交互式Agent基准测试家族，与SWE-bench、WebArena等评测共同构成AI Agent能力的多维评估体系74% similarUnverifiedIAB（Intelligent Agent Benchmarking）workshop聚焦智能体评估70% similarUnverifiedSWE-bench的任务设计本质上是单次任务，Agent在干净环境中处理独立Issue，掩盖了跨会话记忆缺失的问题68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/830340API
curl https://kongchang.com/api/v1/knowledge/claims/830340MCP
get_claim(id=830340)