1 paper
Kai Yu, Zhenhao Zhou, Junhao Zeng +8
Repository-level issue resolution benchmarks have become a standard testbed for evaluating LLM-based agents, yet success is still predominantly measured by test pass rates. In prac…