Spatial Reasoning on 3DSR (Component Metrics)
68.6Height ScoreOurs
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Ours2026.04 | 68.6 | 84.8 | 54.9 | 67.1 | 70.7 | |
| Gemini-2.5-FlashModel Type=Closed-source Models2026.04 | 68.3 | 78.7 | 52.8 | 57.6 | 65.1 | |
| GPT-5.2Model Type=Closed-source Models, Scale=~100B active parameters, Training Data=≥30T tokens2026.04 | 64 | 73 | 57.7 | 57.4 | 63.5 | |
| InternVL3.5-8BModel Type=Open-source Models2026.04 | 59.7 | 72.7 | 48.4 | 51.2 | 58.9 | |
| Qwen3-VL-8BModel Type=Open-source Models2026.04 | 56.7 | 72.4 | 49.3 | 54.1 | 60.4 | |
| Cambrian-1-8BModel Type=Open-source Models2026.04 | 54.9 | 64.2 | 43.4 | 52.3 | 54.9 | |
| SpaceQwen2.5-VL-3BModel Type=Spatially Specialized Models2026.04 | 54.3 | 59.8 | 42.9 | 47.4 | 51.5 | |
| LLaVA-OneVision-1.5-8BModel Type=Open-source Models2026.04 | 52.9 | 60.5 | 43.2 | 47.4 | 51.7 | |
| Spatial-MLLM-7BModel Type=Spatially Specialized Models, Evaluation Mode=Exclusively processes video inputs (single-frame videos)2026.04 | 52.9 | 49.3 | 42.7 | 46.6 | 47.5 | |
| LLaVa-1.6-Mistral-7BModel Type=Open-source Models2026.04 | 51.1 | 57.5 | 40.2 | 46.3 | 49.5 | |
| SpatialLLM-8BModel Type=Spatially Specialized Models2026.04 | 50.6 | 76.7 | 49.5 | 48.5 | 58.4 | |
| VLM3R-7BModel Type=Spatially Specialized Models, Evaluation Mode=Exclusively processes video inputs (single-frame videos)2026.04 | 46 | 42.3 | 31.1 | 26.5 | 35.3 | |
| SpaceMantis-8BModel Type=Spatially Specialized Models2026.04 | 41.4 | 57.1 | 42.9 | 47.2 | 48.9 |