Datasets:
Download tests/test_stats.py from Nabidnur/RoboRACE-Bench: direct link, hf CLI and curl.
- Browser
- Download file 836 Bytes
-
https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/tests/test_stats.py
- Command line
-
hf download hf://datasets/Nabidnur/RoboRACE-Bench/tests/test_stats.py
-
curl -L -o test_stats.py https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/tests/test_stats.py
836 Bytes
| import json | |
| from robotics.eval.stats import compare, main, wilson | |
| def test_wilson_known_value(): | |
| lo, hi = wilson(45, 50) | |
| assert abs(lo - 0.786) < 0.005 and abs(hi - 0.957) < 0.005 | |
| def test_wilson_edges(): | |
| assert wilson(0, 0) == (0.0, 1.0) | |
| lo, hi = wilson(50, 50) | |
| assert hi == 1.0 and lo > 0.9 | |
| def test_compare_significance(): | |
| assert compare(30, 100, 50, 100)["p_value"] < 0.01 | |
| assert compare(50, 100, 52, 100)["p_value"] > 0.5 | |
| def test_cli(tmp_path, capsys): | |
| f = tmp_path / "r.jsonl" | |
| f.write_text("\n".join(json.dumps({"run": r, "suite": "s", "success": i < k}) | |
| for r, k in [("A", 3), ("B", 8)] for i in range(10))) | |
| main([str(f), "--compare", "A", "B"]) | |
| out = capsys.readouterr().out | |
| assert "30.0%" in out and "80.0%" in out and "B - A: +50.0%" in out | |