1 paper · 1 filter
Siyu Zhu, Anastasiya Karpovich, Albert Chen +6
We tackle the challenge of training reliable code-fixing agents in real repositories, where complex builds and shifting dependencies make evaluation unstable. We developed a verifi…