Visual Perception on V* (Component Metrics)
90Overall ScoreDeepEyes
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DeepEyesbackbone=Qwen2.5-VL-7B2026.02 | 90 | 92.1 | 86.8 | |
| CapImaginebackbone=Qwen2.5-VL-7B2026.02 | 85.9 | 87.8 | 82.9 | |
| Monetbackbone=Qwen2.5-VL-7B, reported_from_prior_work=true2026.02 | 83.3 | 83.5 | 82.9 | |
| CapImaginebackbone=Qwen2.5-VL-7B, rewriting=false2026.02 | 82.7 | 87.8 | 77.6 | |
| CapImaginebackbone=Qwen2.5-VL-7B, filtering=false2026.02 | 82.7 | 82.6 | 82.9 | |
| LVRbackbone=Qwen2.5-VL-7B2026.02 | 81.7 | 84.4 | 77.6 | |
| PixelReasonerbackbone=Qwen2.5-VL-7B2026.02 | 80.6 | 83.5 | 76.3 | |
| Monetbackbone=Qwen2.5-VL-7B, subset=true2026.02 | 79.6 | 82.6 | 75 | |
| Qwen2.5VL-7Bbackbone=Qwen2.5VL-7B2026.02 | 76.4 | 77.4 | 75 | |
| InternVL3-8Bbackbone=InternVL3-8B2026.02 | 72.3 | 73 | 71.1 | |
| GPT-4oreported_from_prior_work=true2026.02 | 67.5 | 72.2 | 60.5 |