List your company

Subject Knowledge Evaluation on MMLU

48.7MMLU Overall AccuracyPractical (R̄_FPO)
48.38848.46948.5548.631May 5, 2026

Evaluation results

MethodLinks
Practical (R̄_FPO)Evaluation Protocol=50-question stratified sample per subject2026.0548.741.25453.150.6
Relaxed (R̃_FPO)Evaluation Protocol=50-question stratified sample per subject2026.0548.641.554.352.649.5
Standard RLHFEvaluation Protocol=50-question stratified sample per subject2026.0548.441.353.752.550