SWE-agent: Princeton's Open-Source AI for Software Engineering

Princeton University open-sourced SWE-agent, a framework that enables language models to autonomously solve real GitHub issues. SWE-agent achieved a 12.5% resolution rate on the SWE-bench benchmark, the highest among open-source solutions at the time. It introduces the concept of Agent-Computer Interface (ACI), optimizing the interaction between the model and the code environment.