TestEvo-Bench Evaluates Test and Code Co-Evolution via Real Commit Histories
July 3, 2026
TestEvo-Bench provides an executable benchmark for evaluating how automation agents handle code changes. It uses real commit histories to test two tracks: generating new tests for new behaviors and updating existing failing tests to match modified software behavior.
HOW THIS AFFECTS YOU
●
builderYou can use this to validate whether your coding agents actually produce runnable, semantically correct tests.
●
researcherThis allows you to move beyond static metadata to execution-grounded metrics for co-evolution tasks.