swe-sweep is a new benchmarking framework by Meta designed to evaluate language models on finding and fixing bugs. The tool tests how effectively AI agents navigate and modify large-scale, real-world GitHub repositories. This helps developers measure the true coding capabilities of models beyond isolated snippets.
Opening Kapyn…