Spatial Reasoning on SPBench MV
81.5NQ ScoreOneCanvas
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OneCanvasEvaluation Protocol=Zero-shot evaluation2026.06 | 81.5 | 62.8 | 62.7 | |
| SpaceMindEvaluation Protocol=Zero-shot evaluation2026.06 | 73.8 | 66.3 | 53.2 | |
| SpatialLadder-3BBase Model=Qwen2.5-VL-3B2026.05 | 68.2 | 75 | 70.9 | |
| Spatial-MLLM-4BBase Model=Qwen2.5-VL-3B2026.05 | 63.7 | 58.9 | 53.1 | |
| SpaceR-7BBase Model=Qwen2.5-VL-7B2026.05 | 63.2 | 53.7 | 59.4 | |
| Spatial-MLLM-4BEvaluation Protocol=Zero-shot evaluation2026.06 | 61.3 | 38.1 | 49.3 | |
| CaMo-3BBase Model=Qwen2.5-VL-3B2026.05 | 61.2 | 72.8 | 65.8 | |
| SpaceR-7BEvaluation Protocol=Zero-shot evaluation2026.06 | 58.5 | 35.7 | 61.5 | |
| VILASR-7BEvaluation Protocol=Zero-shot evaluation2026.06 | 57.9 | 36.6 | 63.7 | |
| VILASR-7BBase Model=Qwen2.5-VL-7B2026.05 | 56.2 | 59.6 | 57.6 | |
| Gemini-2.0-FlashModel Type=Proprietary2026.05 | 51.9 | 50.7 | 56.5 | |
| Video-R1Evaluation Protocol=Zero-shot evaluation2026.06 | 42.8 | 27.7 | 62 | |
| GPT-4oModel Type=Proprietary2026.05 | 40.7 | 59.4 | 48.2 | |
| InternVL-2.5-4BModel Type=Open-Source2026.05 | 37.7 | 51.4 | 43.2 | |
| InternVL-2.5-8BModel Type=Open-Source2026.05 | 37.3 | 47.5 | 41.4 | |
| Video-R1Base Model=Qwen2.5-VL-7B2026.05 | 32.5 | 53 | 40.7 | |
| Qwen2.5-VL-7BBase Model=Qwen2.5-VL-7B2026.05 | 28.9 | 49.8 | 37.3 | |
| Qwen2.5-VL-3BBase Model=Qwen2.5-VL-3B2026.05 | 25.6 | 53.2 | 36.6 | |
| Kimi-VL-A3BModel Type=Open-Source2026.05 | 23.3 | 57.6 | 37 | |
| LLaVA-OneVision-7BModel Type=Open-Source2026.05 | 20.6 | 49.6 | 32.2 |