Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?
This work evaluates coding agents' completion performance in two complementary settings: established SWE-bench tasks from popular repositories, with LLM-generated context files, and a novel collection of issues from repositories containing developer-committed context files.
Thibaud Gloaguen, Niels Mündler, Mark Niklas Müller et al.
· arXiv.org · 28 citations
· ⚡1