Solid Geometry Reasoning on Mathverse solid (test)
92.1Overall AverageHuman
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HumanModel Category=Human Performances2026.05 | 92.1 | 93 | 89.4 | 97.8 | 88.5 | |
| Hilbert-Geo (Ground-Truth)Model Category=Hilbert-Geo, Number of samples=452026.05 | 86.3 | 87.5 | 87.1 | 86.9 | 81.1 | |
| Hilbert-Geo (Gemini-2.5-Pro)Model Category=Hilbert-Geo, Number of samples=452026.05 | 84.1 | 85.9 | 84.9 | 80.2 | 78.3 | |
| GPT-5Model Category=Closed-source MLLMs2026.05 | 62.9 | 64.9 | 67.1 | 52.9 | 56.1 | |
| Gemini 2.5 ProModel Category=Closed-source MLLMs2026.05 | 59.7 | 60.6 | 61.9 | 54.1 | 60.1 | |
| Claude 3.7 SonnetModel Category=Closed-source MLLMs2026.05 | 54.7 | 57.2 | 55.3 | 50.6 | 52.4 | |
| Gemini 2.5 FlashModel Category=Closed-source MLLMs2026.05 | 47.5 | 50.1 | 47.2 | 44.5 | 45.9 | |
| GPT-4oModel Category=Closed-source MLLMs2026.05 | 42.4 | 45.1 | 36.6 | 44.2 | 46.7 | |
| Llama 3.3 70BModel Category=Open-source MLLMs2026.05 | 36 | 38.2 | 36.5 | 33.8 | 33.1 | |
| Deepseek-V3 67BModel Category=Open-source MLLMs2026.05 | 35.4 | 39.1 | 33.7 | 34.7 | 32.2 | |
| Qwen2.5-VL-Instruct-32BModel Category=Open-source MLLMs2026.05 | 31 | 33.9 | 29.1 | 29.8 | 30.4 | |
| Qwen2.5-VL-Instruct-7BModel Category=Open-source MLLMs2026.05 | 22.9 | 24.7 | 20.4 | 23.5 | 24.2 |