Spatial VQA on CV-Bench-3D Level 2
97.58Accuracy (%)HiSpatial-3B
Evaluation results
| Method | Links | |
|---|---|---|
| HiSpatial-3BInput=RGB-XYZ2026.03 | 97.58 | |
| RoboRefer-8B-SFTInput=RGB-D2026.03 | 95.92 | |
| HiSpatial-3B-RGBInput=RGB2026.03 | 95.58 | |
| Qwen-3-VL-30B-A3BInput=RGB2026.03 | 92 | |
| Gemini-2.5-ProInput=RGB2026.03 | 90.8 | |
| Qwen-3-VL-8BInput=RGB2026.03 | 90.66 | |
| SpatialRGPT-8BInput=RGB-D2026.03 | 89.15 | |
| InternVL3.5-14BInput=RGB2026.03 | 86.75 | |
| InternVL3.5-8BInput=RGB2026.03 | 85.09 | |
| Claude-3.7-SonnetInput=RGB2026.03 | 85 | |
| GPT-4oInput=RGB2026.03 | 84.9 | |
| SpatialBot-3BInput=RGB-D2026.03 | 69.08 | |
| SpaceLLaVA-13BInput=RGB2026.03 | 68.5 | |
| PaliGemma2-3BInput=RGB2026.03 | 42.17 |