1 paper · 1 filter
Vlad Sobal, Shuo Yang, Yuting Zhang +2
We introduce StaminaBench, a benchmark that measures the stamina of coding agents: how many consecutive interaction turns (change requests) they can handle before failing. Unlike t…