1 paper · 1 filter
Wuya Chen, Yihao yang, Yang Cao +1
Modern LLM coding agents such as Claude Code and OpenHands share a common inefficiency: they spend much of their token budget finding the file to patch, rather than patching it. On…