Hot eventLive
TestGRAD论文:SWE-Agent集成测试套件优化框架
1 reports1 sources4 hr ago updated
Get the story
AI overview
2026年10月8日,arXiv软件工程栏目发布一手论文,介绍TestGRAD框架。该框架面向SWE-agent集成,用于自动优化测试套件,把测试生成过程建模为测试空间优化。论文提出三个机制:差分损失(differential loss)、全CRUD梯度与失败模式动量(Failure Pattern Momentum),目标是让测试套件能够区分候选补丁,并避免在无效方向上重复生成。目前报道仅涉及框架机制与目标,未披露实验数据或后续验证进展。
Generated from reports · updated 3 hr ago
Timeline
Follow the coverage from different angles.
Oct 8, 2026
- arXiv · Software EngineeringTestGRAD:用失败模式动量演化测试套件,优化 SWE-Agent 集成
TestGRAD 是一个面向 SWE-agent 集成的自动测试优化框架,把测试生成过程建模为测试空间优化。它提出差分损失(differential loss)、全 CRUD 梯度与失败模式动量(Failure Pattern Momentum)三个机制,让测试套件能区分候选补丁并避免重复无效方向。
Heat trend
Current heat 9·Comparable peak 10(Oct 8)·Comparable change over 24 hours –
The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.