Multimodal Capability Evaluation on MMStar
68.3Overall ScoreBaseline
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| BaselineBackbone=InternVL3-8B, Retained Tokens=All 1280 Tokens, Reduction Percentage=100%2026.06 | 68.3 | — | — | — | — | — | — | 74.9 | 100 | |
| PStarBackbone=Qwen2.5-VL-7B2026.05 | 68 | 76.8 | 61.2 | 70.8 | 73.2 | 72.4 | 53.6 | — | — | |
| PStarBackbone=Qwen2-VL-7B2026.05 | 65.7 | 74.8 | 60 | 69.6 | 66.8 | 67.6 | 55.6 | — | — | |
| BaselineBackbone=Qwen2.5-VL-7B-Instruct, Retained Tokens=All 1296 Tokens, Reduction Percentage=100%2026.06 | 65.5 | — | — | — | — | — | — | 73.1 | 100 | |
| Qwen2.5-VL-7BModel=Qwen2.5-VL-7B*2026.05 | 64.1 | 72 | 57.6 | 70.4 | 67.6 | 67.2 | 49.6 | — | — | |
| TOPSBackbone=InternVL3-8B, Retained Tokens=256 Tokens, Reduction Percentage=↓ 80.0%2026.06 | 62.8 | — | — | — | — | — | — | 72.4 | 96.7 | |
| InternVL2-8BModel=InternVL2-8B*2026.05 | 61.5 | 70.4 | 48.8 | 68 | 63.2 | 67.2 | 51.6 | — | — | |
| FastVBackbone=InternVL3-8B, Retained Tokens=256 Tokens, Reduction Percentage=↓ 80.0%2026.06 | 61.2 | — | — | — | — | — | — | 71.2 | 95.1 | |
| Qwen2-VL-7BModel=Qwen2-VL-7B*2026.05 | 60.7 | 72 | 56.4 | 68 | 61.6 | 58.4 | 48 | — | — | |
| TOPSBackbone=Qwen2.5-VL-7B-Instruct, Retained Tokens=256 Tokens, Reduction Percentage=↓ 80.2%2026.06 | 59.5 | — | — | — | — | — | — | 70.4 | 96.3 | |
| DivPruneBackbone=InternVL3-8B, Retained Tokens=256 Tokens, Reduction Percentage=↓ 80.0%2026.06 | 58.9 | — | — | — | — | — | — | 69.3 | 92.5 | |
| DivPruneBackbone=Qwen2.5-VL-7B-Instruct, Retained Tokens=256 Tokens, Reduction Percentage=↓ 80.2%2026.06 | 57.9 | — | — | — | — | — | — | 68.8 | 94.1 | |
| CDPrunerBackbone=Qwen2.5-VL-7B-Instruct, Retained Tokens=256 Tokens, Reduction Percentage=↓ 80.2%2026.06 | 57.7 | — | — | — | — | — | — | 68.9 | 94.3 | |
| TOPSBackbone=InternVL3-8B, Retained Tokens=128 Tokens, Reduction Percentage=↓ 90.0%2026.06 | 57.1 | — | — | — | — | — | — | 68.8 | 91.9 | |
| GPT-4V (0409)Model=GPT-4V (0409)*2026.05 | 56 | 69.6 | 44.8 | 67.2 | 59.2 | 60.8 | 34.4 | — | — | |
| CDPrunerBackbone=InternVL3-8B, Retained Tokens=256 Tokens, Reduction Percentage=↓ 80.0%2026.06 | 55.7 | — | — | — | — | — | — | 67.4 | 90 | |
| FastVBackbone=Qwen2.5-VL-7B-Instruct, Retained Tokens=256 Tokens, Reduction Percentage=↓ 80.2%2026.06 | 55.5 | — | — | — | — | — | — | 67.8 | 92.7 | |
| VisionZipBackbone=InternVL3-8B, Retained Tokens=256 Tokens, Reduction Percentage=↓ 80.0%2026.06 | 55.1 | — | — | — | — | — | — | 67.8 | 90.5 | |
| TOPSBackbone=Qwen2.5-VL-7B-Instruct, Retained Tokens=128 Tokens, Reduction Percentage=↓ 90.1%2026.06 | 53.9 | — | — | — | — | — | — | 67.9 | 92.9 | |
| CDPrunerBackbone=Qwen2.5-VL-7B-Instruct, Retained Tokens=128 Tokens, Reduction Percentage=↓ 90.1%2026.06 | 53.8 | — | — | — | — | — | — | 65.5 | 89.6 | |
| DivPruneBackbone=Qwen2.5-VL-7B-Instruct, Retained Tokens=128 Tokens, Reduction Percentage=↓ 90.1%2026.06 | 52.3 | — | — | — | — | — | — | 65.5 | 89.6 | |
| DivPruneBackbone=InternVL3-8B, Retained Tokens=128 Tokens, Reduction Percentage=↓ 90.0%2026.06 | 52 | — | — | — | — | — | — | 64.6 | 86.2 | |
| CDPrunerBackbone=InternVL3-8B, Retained Tokens=128 Tokens, Reduction Percentage=↓ 90.0%2026.06 | 51.5 | — | — | — | — | — | — | 62.8 | 83.8 | |
| PStarBackbone=Qwen2-VL-2B2026.05 | 50.4 | 55.6 | 35.6 | 61.6 | 54.4 | 56.4 | 38.8 | — | — | |
| InternVL2-2BModel=InternVL2-2B*2026.05 | 49.8 | 64 | 38.8 | 60.8 | 42.8 | 49.2 | 43.2 | — | — | |
| LLaMA-3.2-11B-VisionModel=LLaMA-3.2-11B-Vision*2026.05 | 49.8 | 66 | 46.4 | 57.6 | 50.8 | 45.2 | 32.8 | — | — | |
| VisionZipBackbone=InternVL3-8B, Retained Tokens=128 Tokens, Reduction Percentage=↓ 90.0%2026.06 | 49.2 | — | — | — | — | — | — | 60.3 | 80.5 | |
| Qwen2-VL-2BModel=Qwen2-VL-2B*2026.05 | 47.5 | 60 | 50 | 58.4 | 43.2 | 40.8 | 32.8 | — | — | |
| FastVBackbone=InternVL3-8B, Retained Tokens=128 Tokens, Reduction Percentage=↓ 90.0%2026.06 | 47.3 | — | — | — | — | — | — | 60.4 | 80.6 | |
| FastVBackbone=Qwen2.5-VL-7B-Instruct, Retained Tokens=128 Tokens, Reduction Percentage=↓ 90.1%2026.06 | 43.7 | — | — | — | — | — | — | 57 | 78 | |
| LLaVA-v1.5-7BModel=LLaVA-v1.5-7B*2026.05 | 33.1 | 57.2 | 24.4 | 41.6 | 28.4 | 26.4 | 20.4 | — | — |