ScalingIntelligence/swe-bench-verified-codebase-content

SWE-Bench Verified Codebase Content Dataset Introduction SWE-bench is a popular benchmark that measures how well systems can solve real-world software engineering problems. To solve SWE-bench problems, systems need to interact with large codebases that have long commit histories. Interacting with these codebases in an agent loop using git naively can be slow, and the repositories themselves take up large amounts of storage space. This dataset provides the… See the full description on the dataset page: https://huggingface.co/datasets/ScalingIntelligence/swe-bench-verified-codebase-content.