SWE-bench/experiments

Open sourced predictions, execution logs, trajectories, and results from model inference + evaluation runs on the SWE-bench task.