Python: Remove duplicate samples (#3899)

* Remove duplicate samples

* Correct paths

* Update readme

* Update readme

* Fix ruff

---------

Co-authored-by: Evan Mattson <35585003+moonbox3@users.noreply.github.com>
This commit is contained in:
Tao Chen
2026-02-12 15:46:41 -08:00
committed by GitHub
Unverified
parent 1441fd903c
commit e064f943ae
25 changed files with 173 additions and 2772 deletions
@@ -186,7 +186,7 @@ def gaia_scorer(model_answer: str, ground_truth: str) -> bool:
if is_float(ground_truth):
# numeric exact match after normalization
return _normalize_number_str(model_answer) == float(ground_truth)
return abs(_normalize_number_str(model_answer) - float(ground_truth)) < 1e-6
if any(ch in ground_truth for ch in [",", ";"]):
# list with per-element compare (number or string)
gt_elems = _split_string(ground_truth)
@@ -196,7 +196,7 @@ def gaia_scorer(model_answer: str, ground_truth: str) -> bool:
comparisons = []
for ma, gt in zip(ma_elems, gt_elems, strict=False):
if is_float(gt):
comparisons.append(_normalize_number_str(ma) == float(gt))
comparisons.append(abs(_normalize_number_str(ma) - float(gt)) < 1e-6)
else:
comparisons.append(_normalize_str(ma, remove_punct=False) == _normalize_str(gt, remove_punct=False))
return all(comparisons)