Rationale Support Quality and Alignment on MedQA
95.65Accuracy (Q)GRPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GRPOModel=Ministral 3 8B2026.06 | 95.65 | 5.79 | |
| CoRAModel=Ministral 3 8B2026.06 | 94.99 | 5.08 | |
| BaseModel=Ministral 3 8B2026.06 | 94.9 | 6.09 | |
| SFTModel=Ministral 3 8B2026.06 | 94.88 | 5.35 | |
| CoRAModel=Gemma 2 9B2026.06 | 94.47 | 6.02 | |
| GRPOModel=Gemma 2 9B2026.06 | 82.63 | 16.91 | |
| SFTModel=Gemma 2 9B2026.06 | 82.59 | 17.14 | |
| BaseModel=Gemma 2 9B2026.06 | 81.47 | 18.31 | |
| CoRAModel=Qwen2.5 7B2026.06 | 79.87 | 20.29 | |
| SFTModel=Qwen2.5 7B2026.06 | 77.67 | 21.76 | |
| GRPOModel=Qwen2.5 7B2026.06 | 74.77 | 22.31 | |
| BaseModel=Qwen2.5 7B2026.06 | 73.42 | 26.2 |