Geometric Reasoning on Geomverse-109
28.9Accuracy@1VisuoThink
Evaluation Results
| Method | Links | |
|---|---|---|
| VisuoThinkBackbone=GPT-4o2025.04 | 28.9 | |
| VisuoThinkBackbone=Claude-3.5-sonnet2025.04 | 27.8 | |
| VisuoThink w/o rollout searchBackbone=Claude-3.5-sonnet2025.04 | 26.7 | |
| VisuoThinkBackbone=Qwen2-VL-72B-Instruct2025.04 | 25.6 | |
| VisuoThink w/o rollout searchBackbone=GPT-4o2025.04 | 24.4 | |
| VisuoThink w/o rollout searchBackbone=Qwen2-VL-72B-Instruct2025.04 | 19 | |
| VisualSketchpad + Equation SolverBackbone=Claude-3.5-sonnet2025.04 | 17.8 | |
| VisualSketchpadBackbone=Claude-3.5-sonnet2025.04 | 16.7 | |
| CoTBackbone=Claude-3.5-sonnet2025.04 | 14.4 | |
| VisualSketchpad + Equation SolverBackbone=GPT-4o2025.04 | 13.3 | |
| CoTBackbone=GPT-4o2025.04 | 11.1 | |
| VisualSketchpad + Equation SolverBackbone=Qwen2-VL-72B-Instruct2025.04 | 11.1 | |
| VisualSketchpadBackbone=GPT-4o2025.04 | 8.9 | |
| VisualSketchpadBackbone=Qwen2-VL-72B-Instruct2025.04 | 6.7 | |
| CoTBackbone=Qwen2-VL-72B-Instruct2025.04 | 5.6 |