1 paper
Tue Le, Minh V. T. Thai, Dung Nguyen Manh +2
Existing benchmarks for AI coding agents focus on isolated, single-issue tasks such as fixing a bug or adding a small feature. However, real-world software engineering is a long-ho…