Read the original at HF Daily Papers
AgentSWE is a benchmark that tests coding agents on creating, editing, and optimizing software agents. It includes 25 tasks designed to verify whether built agents meet specific user requirements.
Carried by: HF Daily Papers. First seen: .