Quantitative Spatial Visual Question Answering on SpatialRGPT Quantitative
70.68WidthHiSpatial-3B*
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| HiSpatial-3B*Input=RGB-XYZ (GT)2026.03 | 70.68 | 84.21 | 83.11 | 90.16 | 79.25 | 81.46 | |
| HiSpatial-3B-RGBInput=RGB2026.03 | 69.92 | 84.96 | 66.89 | 71.34 | 68.87 | 72.43 | |
| HiSpatial-3BInput=RGB-XYZ2026.03 | 69.17 | 84.96 | 79.73 | 86.89 | 75.47 | 79.28 | |
| GPT-5Input=RGB2026.03 | 59.09 | 67.66 | 29.72 | 25.21 | 13.72 | 40.47 | |
| MM-Spatial-3BInput=RGB-D2026.03 | 55.6 | 83.5 | 59.5 | 82 | 63.2 | 68.7 | |
| GPT-4oInput=RGB2026.03 | 51.1 | 68.4 | 29.7 | 25.4 | 33 | 41.5 | |
| SpatialRGPT-8BInput=RGB-D2026.03 | 48.9 | 61.7 | 45.9 | 68 | 56.6 | 56.22 | |
| Gemini-2.5-ProInput=RGB2026.03 | 37.2 | 43.51 | 16.78 | 17.39 | 15.53 | 26.57 | |
| InternVL3.5-8BInput=RGB2026.03 | 27.81 | 41.35 | 31.08 | 31.96 | 17.92 | 30.53 | |
| Qwen3VL-8BInput=RGB2026.03 | 27.06 | 40.6 | 31.1 | 31.96 | 18.86 | 30.37 | |
| PaliGemma2-3BInput=RGB2026.03 | 20.3 | 27.07 | 20.27 | 15.57 | 22.64 | 21.18 | |
| RoboRefer-8B-SFTInput=RGB-D2026.03 | 17.29 | 60.15 | 49.32 | 37.7 | 28.3 | 39.25 |