Fine-grained Visual Reasoning on V* OOD
90Overall ScoreDeepEyes
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DeepEyes2026.07 | 90 | 92.1 | 86.8 | |
| TikArtBase Model=Qwen3-VL-8B2026.07 | 89.5 | 91.3 | 86.8 | |
| SIEVEBase Model=Qwen3-VL-8B2026.07 | 88 | 88.7 | 86.8 | |
| TreeVGR2026.07 | 87.4 | 89.5 | 84.2 | |
| MIRROR2026.07 | 83.8 | — | — | |
| BUS-GRPOBase Model=Qwen3-VL-8B2026.07 | 83.8 | 85.2 | 81.6 | |
| BUS-SFTBase Model=Qwen3-VL-8B2026.07 | 82.7 | 83.5 | 81.6 | |
| Thyme2026.07 | 82.2 | 83.5 | 80.3 | |
| DeepEyesV22026.07 | 81.8 | 81.7 | 80.3 | |
| V-Reflection2026.07 | 81.7 | 83.5 | 78.9 | |
| PixelReasoner2026.07 | 80.6 | 83.5 | 76.3 | |
| Gemini-2.5-flash2026.07 | 80.1 | 84.4 | 73.7 | |
| Gemini-2.5-pro2026.07 | 79.1 | 84.4 | 71.1 | |
| Qwen3-VL-8B2026.07 | 77.5 | 80.2 | 73.7 | |
| GPT-52026.07 | 76.4 | 74.8 | 79 | |
| Qwen3-VL-4B2026.07 | 74.9 | 78.3 | 69.7 | |
| Qwen2.5-VL-7B2026.07 | 74.3 | 77.4 | 69.7 | |
| InternVL3-8B2026.07 | 72.3 | 73 | 71.1 | |
| GPT-4o2026.07 | 70.7 | 71.3 | 69.7 | |
| GPT-5-nano2026.07 | 63.9 | 58.3 | 72.4 |