Fine-grained Visual Reasoning on MME-RW-Lite ID
57Overall ScoreTikArt
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TikArtBase Model=Qwen3-VL-8B2026.07 | 57 | 60.5 | 53.5 | |
| SIEVEBase Model=Qwen3-VL-8B2026.07 | 56.4 | 58.4 | 53.2 | |
| GPT-52026.07 | 56.2 | 56.9 | 55.5 | |
| TreeVGR2026.07 | 54.9 | 58.2 | 49.7 | |
| BUS-GRPOBase Model=Qwen3-VL-8B2026.07 | 54.4 | 57.6 | 49.5 | |
| Thyme2026.07 | 54.4 | 58.2 | 48.7 | |
| BUS-SFTBase Model=Qwen3-VL-8B2026.07 | 54.2 | 58.1 | 48.1 | |
| V-Reflection2026.07 | 53.9 | 58.5 | 45 | |
| Gemini-2.5-pro2026.07 | 53.5 | 52.1 | 50.7 | |
| DeepEyes2026.07 | 53.2 | 55.4 | 46.8 | |
| DeepEyesV22026.07 | 52.4 | 54.8 | 48 | |
| Gemini-2.5-flash2026.07 | 52.1 | 49.7 | 47.3 | |
| MIRROR2026.07 | 51.5 | — | — | |
| PixelReasoner2026.07 | 49.7 | 53 | 45.6 | |
| Qwen3-VL-8B2026.07 | 48.6 | 54 | 40.4 | |
| InternVL3-8B2026.07 | 47.9 | 51.1 | 42.9 | |
| Qwen3-VL-4B2026.07 | 47.1 | 51.8 | 39.7 | |
| GPT-4o2026.07 | 45.7 | 49.8 | 53.9 | |
| GPT-5-nano2026.07 | 42.8 | 46.1 | 49.4 | |
| Qwen2.5-VL-7B2026.07 | 42.3 | 46.5 | 35.9 |