Multimodal Reasoning on Vision R1 (test)
58.75AccuracyGRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| GRPOModel Scale=8B, Distillation Strategy=GRPO, Visual Cues (ViCuR)=False2026.06 | 58.75 | |
| OPD+ViCuRModel Scale=8B, Distillation Strategy=OPD, Visual Cues (ViCuR)=True2026.06 | 52.07 | |
| OPDModel Scale=8B, Distillation Strategy=OPD, Visual Cues (ViCuR)=False2026.06 | 51.09 | |
| OPSD+ViCuRModel Scale=8B, Distillation Strategy=OPSD, Visual Cues (ViCuR)=True2026.06 | 45 | |
| OPSDModel Scale=8B, Distillation Strategy=OPSD, Visual Cues (ViCuR)=False2026.06 | 42.2 | |
| Base ModelModel Scale=8B, Distillation Strategy=None, Visual Cues (ViCuR)=False2026.06 | 41.22 | |
| GRPOModel Scale=2B, Distillation Strategy=GRPO, Visual Cues (ViCuR)=False2026.06 | 28.17 | |
| OPD+ViCuRModel Scale=2B, Distillation Strategy=OPD, Visual Cues (ViCuR)=True2026.06 | 26.83 | |
| OPSD+ViCuRModel Scale=2B, Distillation Strategy=OPSD, Visual Cues (ViCuR)=True2026.06 | 26.58 | |
| OPDModel Scale=2B, Distillation Strategy=OPD, Visual Cues (ViCuR)=False2026.06 | 26.34 | |
| OPSDModel Scale=2B, Distillation Strategy=OPSD, Visual Cues (ViCuR)=False2026.06 | 25 | |
| Base ModelModel Scale=2B, Distillation Strategy=None, Visual Cues (ViCuR)=False2026.06 | 23.9 |