# Official SWE-bench grader logs (test output, run logs) are large and reproducible; the
# per-run report JSON and predictions next to them are kept.
_eval/*/logs/
