SCIENCE UNDER AUTOMATION
CritPt independent leaderboard
Snapshot lists leading score 32.9%, Muse Wonder Full (Max). Benchmark explicitly under review.
What this result supports
Excluded from passing quantitative validity gates while under review; no replacement for corrected scores.
Evidence record: under_review. Checked: 2026-10-20.
Where this evidence is used
Theoretical physics
- Set up the physical model: CritPt probes specified problems; the current evaluation is under review.
- Derive or compute a solution: Reported success is partial and benchmark validity is under review.
- Check consistency and correctness: Graded answers do not establish universally reliable derivation checks.
- Choose a productive new explanation: Problem solving does not evaluate autonomous research direction.
- Connect theory to discriminating predictions: No matched broad evaluation in this source set.
- Resolve consistency with observations: Whole research programmes are outside scope.