Multiple Choice Question Answering on MMLU STEM
84.91ScoreQwen2.5(14B)T
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5(14B)TEvaluation Protocol=Generation-based2026.07 | 84.91 | — | — | |
| Qwen2.5(14B)TEvaluation Protocol=Likelihood-based2026.07 | 75.55 | — | — | |
| RG-OPDEvaluation Protocol=Generation-based2026.07 | 59.46 | 2.75 | 2 | |
| Qwen2.5(1.5B)SEvaluation Protocol=Generation-based2026.07 | 56.71 | 0 | 2.4 | |
| Reverse KLEvaluation Protocol=Generation-based2026.07 | 56.3 | -0.41 | 3 | |
| RG-OPDEvaluation Protocol=Likelihood-based2026.07 | 53.85 | 0.28 | 2 | |
| Qwen2.5(1.5B)SEvaluation Protocol=Likelihood-based2026.07 | 53.57 | 0 | 2.4 | |
| TSD-KDEvaluation Protocol=Likelihood-based2026.07 | 53.44 | -0.13 | 2.6 | |
| TSD-KDEvaluation Protocol=Generation-based2026.07 | 52.55 | -4.16 | 2.6 | |
| Reverse KLEvaluation Protocol=Likelihood-based2026.07 | 52.17 | -1.4 | 3 |