Skip to content

fix(metrics): compare mathematical equivalence in Inspect math_scorer (#1430) - #1451

Open
pfolex wants to merge 1 commit into
huggingface:mainfrom
pfolex:fix/inspect-math-scorer-equivalence-1430
Open

pfolex wants to merge 1 commit into
huggingface:mainfrom
pfolex:fix/inspect-math-scorer-equivalence-1430

Conversation

@pfolex

@pfolex pfolex commented Oct 24, 2026

Copy link
Copy Markdown

Fixes #1430.

Summary of Changes

In math_scorer() (src/lighteval/metrics/metrics.py), predictions and targets were compared using direct Python list equality on raw extraction results:

value = "C" if extracted_predictions == extracted_gold else "I"

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[BUG] Inspect math scorer rejects equivalent numerical answers by comparing parser fallback strings

1 participant