Spatial Reasoning on MindCube (Detailed Accuracy Metrics)
43.52Overall AccuracyDCRL
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DCRLModel Category=Our Models2026.06 | 43.52 | 59.2 | 43.5 | 37 | |
| Qwen3-VL-8BModel Category=Our Models2026.06 | 40.01 | 53.2 | 41 | 34.33 | |
| GPT-4oModel Category=Proprietary Model2026.06 | 38.81 | 32.65 | 40.17 | 29.16 | |
| RoboBrainModel Category=Spatial Models2026.06 | 37.38 | 35.8 | 38.28 | 29.53 | |
| Idefics3-8B-Llama3Model Category=Open-Weight Multi Image Models2026.06 | 35.86 | 35.15 | 35.94 | 35.49 | |
| LongVA-7BModel Category=Open-Weight Multi Image Models2026.06 | 29.46 | 35.89 | 29.55 | 24.88 | |
| Qwen2.5-VL-7BModel Category=Open-Weight Multi Image Models2026.06 | 29.26 | 38.76 | 29.5 | 21.35 | |
| SpatialVLMModel Category=Spatial Models2026.06 | 22.81 | 37.65 | 21.26 | 29.39 | |
| InternVL2.5-8BModel Category=Open-Weight Multi Image Models2026.06 | 18.68 | 36.45 | 18.2 | 13.11 |