Geometry problem solving on MathVista
87.15BoN@3 AccuracyGeometryZero-7B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GeometryZero-7BModel Size=7B2025.06 | 87.15 | — | |
| Qwen2.5-7B-Instruct + GRPOModel Size=7B2025.06 | 86.23 | — | |
| Qwen2.5-7B-Instruct + ToRLModel Size=7B2025.06 | 83.48 | — | |
| Qwen2.5-3B-Instruct + GRPOModel Size=3B2025.06 | 82.87 | — | |
| GeometryZero-3BModel Size=3B2025.06 | 82.56 | — | |
| Qwen2.5-3B-Instruct + ToRLModel Size=3B2025.06 | 80.34 | — | |
| Qwen2.5-7B-Instruct + SFTModel Size=7B2025.06 | 76.14 | — | |
| Qwen2.5-3B-Instruct + SFTModel Size=3B2025.06 | 73.08 | — | |
| Qwen2.5-7B-InstructModel Size=7B2025.06 | 68.19 | — | |
| Qwen2.5-3B-InstructModel Size=3B2025.06 | 67.88 | — | |
| GeometryZero-1.5BModel Size=1.5B2025.06 | 60.65 | — | |
| Qwen2.5-1.5B-Instruct + GRPOModel Size=1.5B2025.06 | 57.79 | — | |
| Qwen2.5-1.5B-Instruct + ToRLModel Size=1.5B2025.06 | 57.79 | — | |
| GNS-Llava-1.5-7BModel Size=7B2025.06 | 51.4 | — | |
| Qwen2.5-1.5B-InstructModel Size=1.5B2025.06 | 47.7 | — | |
| G-Llava-7BModel Size=7B2025.06 | 46.92 | — | |
| Qwen2.5-1.5B-Instruct + SFTModel Size=1.5B2025.06 | 43.11 | — | |
| GeometryZero-VL-7BTraining=GCPO2025.06 | — | 83.64 | |
| Qwen2.5-VL-7B-InstructBase Model=Qwen2.5-VL-7B-Instruct2025.06 | — | 63.29 | |
| Qwen2.5-VL-7B-Instruct + GRPOTraining=GRPO2025.06 | — | 81.23 | |
| Qwen2.5-VL-7B-Instruct + SFTTraining=SFT2025.06 | — | 71.95 | |
| Qwen2.5-VL-7B-Instruct + ToRLTraining=ToRL2025.06 | — | 78.46 |