SWE-bench is a benchmark for evaluating large language models on real-world software issues...