Type: url
Original: https://arxiv.org/abs/2405.15793
Status: generated
Summary
“SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering” (Yang, Jimenez, Wettig, Lieret, Yao, Narasimhan, Press; arXiv:2405.15793) argues that language model agents are a new category of end users who benefit from interfaces built specifically for them, and introduces SWE-agent, a system that lets LM agents autonomously use computers to solve software engineering tasks. Its custom agent-computer interface (ACI) is credited with enhancing the agent’s ability to create and edit code files, navigate repositories, and execute tests and other programs. The paper reports state-of-the-art pass@1 rates of 12.5% on SWE-bench and 87.7% on HumanEvalFix.