Fine-grained image perception on V*
91.3Attribute ScoreDeepEyes
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DeepEyesPrefilling=Multiple, Cropping=Yes, Tool-using=true2026.06 | 91.3 | 88.2 | 90.1 | |
| VisReflectPrefilling=Single, Tool-using=false2026.06 | 86.7 | 82.3 | 84.9 | |
| DeepEyesAccess=Open-Source2026.07 | 84.35 | 81.58 | 83.25 | |
| AMVLAccess=Open-Source, Model size=7B2026.07 | 84.35 | 84.21 | 84.29 | |
| Thyme-VLPrefilling=Multiple, Cropping=Yes, Tool-using=true2026.06 | 83.5 | 80.3 | 82.2 | |
| Qwen2.5-VL-7BAccess=Open-Source, Model size=7B, Training=SFT2026.07 | 83.48 | 78.95 | 81.68 | |
| MonetAccess=Open-Source2026.07 | 83.48 | 82.89 | 83.25 | |
| DeepEyesPrefilling=Single, Cropping=No, Tool-using=true2026.06 | 82.6 | 76.3 | 80.1 | |
| PixelReasonerPrefilling=Single, Tool-using=true2026.06 | 81.7 | 76.3 | 79.6 | |
| Thyme-VLPrefilling=Single, Cropping=No, Tool-using=true2026.06 | 81.7 | 78.9 | 80.6 | |
| LVRPrefilling=Single, Tool-using=true2026.06 | 81.7 | 79 | 80.6 | |
| Mull-TokensAccess=Open-Source2026.07 | 81.58 | 77.49 | 79.06 | |
| PixelReasonerAccess=Open-Source2026.07 | 80.87 | 81.58 | 81.15 | |
| Qwen2.5-VLPrefilling=Single, Tool-using=false2026.06 | 80.8 | 76.3 | 79.1 | |
| PAPOPrefilling=Single, Tool-using=true2026.06 | 80.8 | 77.6 | 79.5 | |
| LVRAccess=Open-Source2026.07 | 80 | 82.89 | 81.15 | |
| Qwen2.5-VL-7BAccess=Open-Source, Model size=7B, Training=SFT + GRPO2026.07 | 78.26 | 78.95 | 78.53 | |
| Qwen2.5-VL-7BAccess=Open-Source, Model size=7B2026.07 | 77.39 | 75 | 76.44 | |
| GPT-4oAccess=Proprietary2026.07 | 72.2 | 60.5 | 67.5 |