Fine-grained Visual Reasoning on V-Star
95.7Overall Accuracyo3 OpenAI
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| o3 OpenAI2026.07 | 95.7 | — | — | |
| P2R-8BParameters=8B2026.07 | 93.7 | 92.2 | 96.1 | |
| P2R-4BParameters=4B2026.07 | 93.2 | 92.2 | 94.7 | |
| ZoomEye-7BParameters=7B2026.07 | 90.6 | 93.9 | 85.5 | |
| DeepEyes-7BParameters=7B2026.07 | 90.1 | 91.3 | 88.2 | |
| Qwen3-VL-Instruct-32BParameters=32B2026.07 | 86.9 | 86.1 | 88.2 | |
| P2R-2BParameters=2B2026.07 | 84.3 | 84.3 | 84.2 | |
| Qwen3-VL-Instruct-8BParameters=8B2026.07 | 83.8 | 84.3 | 82.9 | |
| Thyme-7BParameters=7B2026.07 | 82.2 | 83.5 | 80.3 | |
| Qwen3-VL-Instruct-4BParameters=4B2026.07 | 81.7 | 79.1 | 85.5 | |
| DyFo-7BParameters=7B2026.07 | 81.2 | 80 | 82.9 | |
| PixelReasoner-7BParameters=7B2026.07 | 80.6 | 83.5 | 76.3 | |
| Qwen3-VL-Instruct-2BParameters=2B2026.07 | 74.9 | 73 | 77.6 | |
| GPT-4o2026.07 | 66 | — | — |