Spatial Reasoning on MindCube tiny (test)
89.5Rot. AccuracyGemini-2.5-Pro
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Gemini-2.5-ProParadigm=Baseline Foundation VLMs2025.11 | 89.5 | 54.5 | 48.8 | 57.5 | |
| SpatialDreamerBackbone=Qwen2.5-VL-7B2025.12 | 87.5 | 93.5 | 78.2 | 84.9 | |
| Qwen3-VL-ThinkingParadigm=Baseline Foundation VLMs2025.11 | 87 | 47.3 | 35 | 47.3 | |
| GCAParadigm=Tool-Integrated Spatial Agents2025.11 | 82 | 61.8 | 59.8 | 64.2 | |
| GLM-4.5VParadigm=Baseline Foundation VLMs2025.11 | 60 | 25.5 | 42.2 | 39.6 | |
| 3DThinkerBackbone=Qwen2.5-VL-7B2025.12 | 55 | 76 | 83 | 76 | |
| Ego3D-VLM2025.12 | 54.3 | 69.5 | 65.5 | 44.4 | |
| Claude-4-Sonnet2025.12 | 48.4 | 47.6 | 44.2 | 44.8 | |
| Qwen2.5-VL-7B2025.12 | 38.7 | 21.4 | 29.5 | 29.3 | |
| Spatial-MLLM2025.12 | 38.4 | 32.8 | 20.9 | 32.1 | |
| Space-Qwen2025.12 | 38 | 26.3 | 33.7 | 33.3 | |
| SpatialVLM2025.12 | 37.7 | 29.3 | 21.3 | 22.8 | |
| Mantis-8B2025.12 | 37.6 | 51 | 40.2 | 41.1 | |
| DeepSeek-VL22025.12 | 37 | 26.9 | 50.4 | 47.6 | |
| LLaVA-OV-7B2025.12 | 36.5 | 44.1 | 48.4 | 47.4 | |
| InternVL2.5-8B2025.12 | 36.5 | 13.1 | 18.2 | 18.7 | |
| LongVA-7B2025.12 | 35.9 | 24.9 | 29.6 | 29.5 | |
| RoboBrain2025.12 | 35.8 | 29.5 | 38.3 | 37.4 | |
| LLaVA-Video-7B2025.12 | 35.7 | 30.1 | 43.6 | 41.9 | |
| VILASRParadigm=Training-based Spatial VLMs2025.11 | 34.4 | 25.7 | 29.4 | 29.1 | |
| SpatialLLMParadigm=Training-based Spatial VLMs2025.11 | 34 | 26.8 | 33 | 31.1 | |
| Spatial-MLLMParadigm=Training-based Spatial VLMs2025.11 | 33.8 | 34.5 | 28.3 | 32.1 | |
| GPT-4oParadigm=Baseline Foundation VLMs2025.11 | 33.5 | 35 | 37.2 | 35.8 | |
| TIGeRParadigm=Tool-Integrated Spatial Agents2025.11 | 33 | 28.3 | 26.7 | 28.3 | |
| GPT-4o2025.12 | 32.6 | 29.2 | 40.2 | 38.8 | |
| VLaserParadigm=Training-based Spatial VLMs2025.11 | 31.5 | 24.8 | 38.2 | 32.6 | |
| SpatialLadderParadigm=Training-based Spatial VLMs2025.11 | 30.5 | 39.8 | 47.8 | 42.3 | |
| Video-R1Paradigm=Training-based Spatial VLMs2025.11 | 30 | 30.5 | 41.3 | 35.8 | |
| SpaceRParadigm=Training-based Spatial VLMs2025.11 | 29.8 | 30 | 26.8 | 28.3 | |
| RoboBrain-2.0Paradigm=Training-based Spatial VLMs2025.11 | 29.7 | 35.8 | 45.2 | 39.6 |