Fine-grained Visual Reasoning on HR-Bench-4K OOD
87.9Overall ScoreGemini-2.5-pro
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Gemini-2.5-pro2026.07 | 87.9 | 88.8 | 87 | |
| TikArtBase Model=Qwen3-VL-8B2026.07 | 82.3 | 93.8 | 70.8 | |
| SIEVEBase Model=Qwen3-VL-8B2026.07 | 81.5 | 90 | 73 | |
| Gemini-2.5-flash2026.07 | 81 | 85 | 77 | |
| BUS-GRPOBase Model=Qwen3-VL-8B2026.07 | 80.1 | 90 | 70.3 | |
| DeepEyesV22026.07 | 77.9 | 92.8 | 63 | |
| GPT-52026.07 | 77.5 | 78 | 77 | |
| TreeVGR2026.07 | 77.1 | 89.5 | 64.8 | |
| Thyme2026.07 | 77 | 91 | 63 | |
| DeepEyes2026.07 | 75.1 | 91.3 | 59 | |
| BUS-SFTBase Model=Qwen3-VL-8B2026.07 | 74.3 | 90 | 58.5 | |
| Qwen3-VL-4B2026.07 | 73.5 | 84.8 | 62.3 | |
| MIRROR2026.07 | 72.9 | — | — | |
| PixelReasoner2026.07 | 72.9 | 86 | 60.3 | |
| V-Reflection2026.07 | 72.6 | 83.5 | 61.8 | |
| Qwen3-VL-8B2026.07 | 72.4 | 88.5 | 56.3 | |
| Qwen2.5-VL-7B2026.07 | 72.1 | 88.8 | 55.5 | |
| InternVL3-8B2026.07 | 70.8 | 79.3 | 62.3 | |
| GPT-5-nano2026.07 | 65.4 | 69 | 61.8 | |
| GPT-4o2026.07 | 63.9 | 68.8 | 59 |