Joint Success Function
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 1
- TestPrism Exposes Coding Agent Test Reliability Gaps
南京大学刘佳恒团队提出测试生成基准 TestPrism,包含来自17个来源的300个测试任务和3000个候选实现。其核心指标 Joint Success Function(JSF)要求测试在原始代码上失败、接受所有正确实现并拒绝所有错误实现;报道显示,14种 Coding Agent 配置中最佳 JSF 仅28.00%。研究还分析了286条失败轨迹,并提出通过异构合成、同伴交叉验证和递归自我改进提升结果的 TestHelix 框架。
36氪 人工智能 · 🔥 0
Experience and discussion from the community
Share my Joint Success Function experienceAsk about Joint Success Function
Nobody has shared their experience with Joint Success Function yet.