SWE-Prime: Fewer Trajectories, Better Performance
Cutting training data in half actually makes AI better at fixing code
Researchers found that training AI models to fix software bugs works better when you use only 10% of successful examples—but pick them carefully. The key is filtering out steps that are inefficient or risky, even when the overall fix works, improving performance by up to 24% compared to using all the data.
Software companies spend enormous resources collecting and labeling training data for AI coding assistants. This work shows they can achieve better results with far less data, cutting costs and training time while actually improving the quality of the AI's solutions. It also reveals a blind spot: more data isn't always better if that data contains bad practices buried inside otherwise successful fixes.