SWE-Exp: Experience-Driven Software Issue Resolution
Silin Chen, Shaoxin Lin, Xiaodong Gu, Yuling Shi, Heng Lian, Longfei Yun, Dong Chen, Weiguo Sun, Lin Cao, Qianxiang Wang
2025-08-04
Summary
This paper talks about SWE-Exp, a system that helps computer programs fix software problems more effectively by learning from past experiences and repairs.
What's the problem?
The problem is that many current software-fixing programs treat each new issue as completely new and don’t use knowledge from previous fixes, leading to repeated trial-and-error and slower problem solving.
What's the solution?
SWE-Exp solves this by creating an experience bank that stores helpful information from past repairs, like understanding the problem and specific code changes. It uses this stored knowledge to guide new software repairs more strategically instead of randomly trying solutions.
Why it matters?
This matters because it makes automated software fixing faster, smarter, and more reliable, which can save a lot of time and effort for developers and improve software quality overall.
Abstract
SWE-Exp enhances software issue resolution by systematically accumulating and leveraging repair expertise from past agent experiences, improving resolution rates.