Multimodal Perception and Reasoning on MMStar
60.7AccuracyFull Model
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Full ModelBackbone=Qwen2-VL-7B, Sparsity Ratio=0%2026.04 | 60.7 | — | 100 | |
| TopoVLMBackbone=Qwen2-VL-7B, Sparsity Ratio=50%2026.04 | 59.1 | — | 95.4 | |
| TAMPBackbone=Qwen2-VL-7B, Sparsity Ratio=50%2026.04 | 57.3 | — | 93.2 | |
| LLM-StreamlineBackbone=Qwen2-VL-7B, Sparsity Ratio=50%2026.04 | 55.8 | — | 91.2 | |
| SparseGPTBackbone=Qwen2-VL-7B, Sparsity Ratio=50%2026.04 | 54.6 | — | 90.1 | |
| ECoFLaPBackbone=Qwen2-VL-7B, Sparsity Ratio=50%2026.04 | 54.1 | — | 90.5 | |
| LLM-PrunerBackbone=Qwen2-VL-7B, Sparsity Ratio=50%2026.04 | 53.2 | — | 88.6 | |
| WandaBackbone=Qwen2-VL-7B, Sparsity Ratio=50%2026.04 | 52.4 | — | 89.3 | |
| OWLBackbone=Qwen2-VL-7B, Sparsity Ratio=50%2026.04 | 51.7 | — | 88.4 | |
| PG-CoTParameters=7B2025.12 | — | 66 | — | |
| Qwen2.5-VLParameters=7B2025.12 | — | 63 | — | |
| Revisual-R1Parameters=7B2025.12 | — | 67 | — | |
| ThinkLite-VLParameters=7B2025.12 | — | 67 | — |