作者位次 #12/16ACL 2026Findings· Findings of ACL
SWE - QA -Pro: A Representative Benchmark and Scalable Training Recipe for Repository-Level Code Understanding
Songcheng Cai · Zhiheng LYU · Yuansheng Ni · Xiangchao Chen · Baichuan Zhou · Shenzhe Zhu · Yi Lu · Haozhe WANG · Chi Ruan · Benjamin Schneider · Weixu Zhang · Xiang Li · Andy Zheng · Yuyu Zhang · Ping Nie · Wenhu Chen
摘要
Agentic repository-level code understanding is essential for automating complex software engineering tasks, yet the field lacks reliable benchmarks. Existing evaluations often overlook the long tail topics and rely on popular repositories where Large Language Models (LLMs) can cheat via memorized...