List your company

Science Question Answering on GPQA Diamond (Accuracy Tracking & Repair)

75.3Initial AccuracyDISC
53.14858.89964.6570.401Jun 19, 2026

Evaluation results

MethodLinks
DISCModel=gpt-5.2, K=32026.0675.379.318101.836.7
CoVeModel=gpt-5.22026.0675.369.717280.6134.7
Self-RefineModel=gpt-5.22026.0675.375.31313126.5
DISCModel=gpt-4.1-nano, K=32026.06545400—0
CoVeModel=gpt-4.1-nano2026.0654539110.829.9
Self-RefineModel=gpt-4.1-nano2026.065454.5321.53.3