Visual Reasoning on Physical Reasoning
9.01J ScoreDense M.
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Dense M.Backbone=Qwen 7B, Pruning Ratio=0%2026.05 | 9.01 | 86.73 | 92.69 | 1.26 | |
| µCRASPBackbone=Qwen 7B, Pruning Ratio=30%2026.05 | 8.87 | 71.3 | 86.64 | 1.74 | |
| Dense M.Backbone=Qwen 3B, Pruning Ratio=0%2026.05 | 8.86 | 85.97 | 92.08 | 1.36 | |
| Dense M.Backbone=Llama 11B, Pruning Ratio=0%2026.05 | 8.85 | 85.8 | 91.82 | 1.44 | |
| Dense M.Backbone=Gemma 4B, Pruning Ratio=0%2026.05 | 8.65 | 83.53 | 92.06 | 1.28 | |
| µCRASPBackbone=Llama 11B, Pruning Ratio=30%2026.05 | 8.55 | 80.7 | 89.72 | 1.8 | |
| Dense M.Backbone=Qwen 2B, Pruning Ratio=0%2026.05 | 8.39 | 78.37 | 90.04 | 1.76 | |
| Attribution P.Backbone=Llama 11B, Pruning Ratio=30%2026.05 | 8.07 | 0 | 85.61 | 2.41 | |
| µCRASPBackbone=Qwen 3B, Pruning Ratio=30%2026.05 | 8.07 | 24.8 | 80.64 | 2.57 | |
| µCRASPBackbone=Gemma 4B, Pruning Ratio=30%2026.05 | 8.02 | 77.3 | 88.71 | 1.9 | |
| Attribution P.Backbone=Gemma 4B, Pruning Ratio=30%2026.05 | 7.72 | 73.4 | 86.18 | 2.12 | |
| µCRASPBackbone=Qwen 2B, Pruning Ratio=30%2026.05 | 7.65 | 63.4 | 86.98 | 2.4 | |
| LLM-PrunerBackbone=Qwen 7B, Pruning Ratio=30%2026.05 | 7.32 | 56.4 | 83.48 | 2.41 | |
| Attribution P.Backbone=Qwen 2B, Pruning Ratio=30%2026.05 | 7.17 | 17.2 | 81.17 | 2.7 | |
| LLM-PrunerBackbone=Qwen 3B, Pruning Ratio=30%2026.05 | 6.73 | 0.1 | 64.61 | 3.49 | |
| Wanda-SPBackbone=Qwen 7B, Pruning Ratio=30%2026.05 | 6.45 | 2.4 | 60.74 | 2.79 | |
| LLM-PrunerBackbone=Qwen 2B, Pruning Ratio=30%2026.05 | 5.35 | 0 | 72.57 | 4.6 | |
| Wanda-SPBackbone=Qwen 3B, Pruning Ratio=30%2026.05 | 5.34 | 0 | 59.96 | 7.47 | |
| Wanda-SPBackbone=Llama 11B, Pruning Ratio=30%2026.05 | 5 | 0 | 60.45 | 3.21 | |
| LLM-PrunerBackbone=Llama 11B, Pruning Ratio=30%2026.05 | 4.69 | 0 | 65.62 | 3.57 | |
| FLAPBackbone=Llama 11B, Pruning Ratio=30%2026.05 | 2.79 | 0 | 56.33 | 5.98 | |
| Wanda-SPBackbone=Gemma 4B, Pruning Ratio=30%2026.05 | 1.92 | 0.1 | 52.85 | 3.5 | |
| Attribution P.Backbone=Qwen 7B, Pruning Ratio=30%2026.05 | 1.33 | 0 | 9.72 | 241 | |
| FLAPBackbone=Qwen 2B, Pruning Ratio=30%2026.05 | 0.9 | 0 | 39.76 | 15.5 | |
| LLM-PrunerBackbone=Gemma 4B, Pruning Ratio=30%2026.05 | 0.78 | 0.1 | 40.37 | 11.9 | |
| FLAPBackbone=Qwen 7B, Pruning Ratio=30%2026.05 | 0.3 | 0 | 41.28 | 41.47 | |
| FLAPBackbone=Gemma 4B, Pruning Ratio=30%2026.05 | 0.3 | 0 | 35.06 | 17.7 | |
| Wanda-SPBackbone=Qwen 2B, Pruning Ratio=30%2026.05 | 0.29 | 0 | 25.6 | 15.6 | |
| FLAPBackbone=Qwen 3B, Pruning Ratio=30%2026.05 | 0.03 | 0 | 32.49 | 66.9 | |
| Attribution P.Backbone=Qwen 3B, Pruning Ratio=30%2026.05 | 0.01 | 0 | 3.77 | — |