Full Eval Dashboard¶
How to Run¶
Run a full evaluation of any benchmark through Harbor. Results appear on this dashboard after committing.
Run all benchmarks (parallel):
Run individual benchmarks:
benchmarks/run-full-eval.sh swebench --solver factory --concurrency 5 --timeout 36000
benchmarks/run-full-eval.sh featurebench --solver factory --concurrency 5 --timeout 36000
benchmarks/run-full-eval.sh terminalbench --solver factory --concurrency 5 --timeout 36000
benchmarks/run-full-eval.sh programbench --solver factory --concurrency 5 --timeout 36000
Commit results:
Use --limit N to run a subset of tasks (useful for testing).
Loading full eval results...