Multimodal Reasoning on R1-Onevision-Bench Physics
34.9AccuracyXMCC
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| XMCCBase Model=Qwen3-VL2026.02 | 34.9 | 114.6 | 3.3 | |
| Prune-on-LogicBase Model=Qwen3-VL2026.02 | 34.4 | 510.3 | 14.8 | |
| No CompressionBase Model=Qwen3-VL2026.02 | 33.1 | 697.1 | 21.1 | |
| StepEntropyBase Model=Qwen3-VL2026.02 | 31.7 | 432.3 | 13.6 | |
| XMCCBase Model=Qwen2.5-VL2026.02 | 28.1 | 116.8 | 4.2 | |
| StepEntropyBase Model=Qwen2.5-VL2026.02 | 26.6 | 396.4 | 14.9 | |
| Prune-on-LogicBase Model=Qwen2.5-VL2026.02 | 25.2 | 513.2 | 20.4 | |
| No CompressionBase Model=Qwen2.5-VL2026.02 | 23.4 | 748 | 32 |