Solid Geometry Reasoning on SolidFGeo2k
81.8Overall Average ScoreHuman
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HumanContext=Human Performances2026.05 | 81.8 | 84.3 | 81.2 | 86.1 | 78.7 | |
| Hilbert-GeoParsing Source=Ground-Truth2026.05 | 78.7 | 80.5 | 84.1 | 76.3 | 75.1 | |
| Hilbert-GeoParsing Source=Gemini-2.5-Pro2026.05 | 77.3 | 80.3 | 79.4 | 76.2 | 74.8 | |
| Gemini 2.5 ProAccess=Closed-source2026.05 | 54.2 | 60.2 | 62.4 | 48.2 | 49.8 | |
| GPT-5Access=Closed-source2026.05 | 50.6 | 51.2 | 55.4 | 40.6 | 46.8 | |
| Claude 3.7 SonnetAccess=Closed-source2026.05 | 47.1 | 39.2 | 40.4 | 52.9 | 56.8 | |
| Gemini 2.5 FlashAccess=Closed-source2026.05 | 39.8 | 50.7 | 55.4 | 34.6 | 36.8 | |
| GPT-4oAccess=Closed-source2026.05 | 35.8 | 25.9 | 27.8 | 39.2 | 40 | |
| Llama 3.3 70BAccess=Open-source2026.05 | 33.6 | 36.3 | 34.4 | 31.1 | 28.7 | |
| Deepseek-V3 67BAccess=Open-source2026.05 | 30.1 | 11.7 | 11.3 | 28.6 | 27.8 | |
| Qwen2.5-VL-Instruct-32BAccess=Open-source2026.05 | 29.8 | 27.2 | 19.9 | 35.2 | 38.1 | |
| Qwen2.5-VL-Instruct-7BAccess=Open-source2026.05 | 20.2 | 17.7 | 10.7 | 25.2 | 30.1 |