Reasoning on MMLU (all subjects, test)
61.5AccuracyMPO
Evaluation results
| Method | Links | |
|---|---|---|
| MPOSolver Model=LLaMA-3 8B-Instruct2026.01 | 61.5 | |
| UntunedSolver Model=LLaMA-3 8B-Instruct2026.01 | 57.21 | |
| TextGradSolver Model=LLaMA-3 8B-Instruct2026.01 | 56.4 | |
| MPOSolver Model=Mistral-7B-Instruct2026.01 | 55.5 | |
| UntunedSolver Model=Mistral-7B-Instruct2026.01 | 53.79 | |
| TextGradSolver Model=Mistral-7B-Instruct2026.01 | 53.7 |