Visually Grounded Reasoning on V* Bench (test)
95Overall Accuracyo3
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| o32026.03 | 95 | — | — | |
| DeepScan-72BBackbone=Qwen2.5-VL-72B, k=∞2026.03 | 94.2 | 94.8 | 93.4 | |
| DeepScan-72BBackbone=Qwen2.5-VL-72B, k=102026.03 | 93.7 | 93.9 | 93.4 | |
| DeepScanBackbone=Qwen2.5-VL-7B, k=∞2026.03 | 91.1 | 93.9 | 86.8 | |
| DeepScanBackbone=Qwen2.5-VL-7B, k=102026.03 | 90.6 | 93 | 86.8 | |
| DeepEyes2026.03 | 90 | 92.1 | 86.8 | |
| ViGoRL2026.03 | 86.4 | — | — | |
| TreeVGRevaluation_protocol=official codes, suffix=†2026.03 | 85.9 | 86.1 | 85.5 | |
| Qwen2.5-VL-72BBackbone=Qwen2.5-VL-72B2026.03 | 84.8 | 90.8 | 80.9 | |
| Dyfoevaluation_protocol=official codes, suffix=†2026.03 | 84.3 | 82.6 | 86.8 | |
| Thyme-VL2026.03 | 82.2 | 83.5 | 80.3 | |
| ZoomRefineevaluation_protocol=official codes, suffix=†2026.03 | 82.2 | 85.3 | 77.6 | |
| PixelReasoner2026.03 | 80.6 | 83.5 | 76.3 | |
| Dyfo-Qvariant=Q2026.03 | 80 | 82.9 | 81.2 | |
| Seal2026.03 | 74.8 | 76.3 | 75.4 | |
| Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2026.03 | 74.3 | 77.4 | 69.7 | |
| Dyfo-Lvariant=L2026.03 | 62.7 | 53.9 | 59.2 |