List your company

Accuracy on GPQA Diamond (Scientific Reasoning)

40.4AccuracyAMARIS
34.78436.24237.739.158May 18, 2026

Evaluation results

MethodLinks
AMARISRubric Strategy=per-instance rubric2026.0540.4
AMARISRubric Strategy=global rubric2026.0539.9
RubricHub (RuRL)RL Training Approach=RuRL2026.0538.8
RuscaRLRL Training Approach=RuscaRL2026.0538.5
RaRRL Training Approach=RaR2026.0537.6
Naive (no RL)RL Training Approach=None2026.0535