Spatial Understanding on MindCube
66.3AccuracyHY-Embodied
Evaluation results
| Method | Links | ||
|---|---|---|---|
| HY-EmbodiedModel Version=0.5 MoT-2B, Thinking Mode=thinking2026.04 | 66.3 | — | |
| Dirichlet-energy regularized LoRAModel=InternVL3-8B2026.05 | 51.7 | — | |
| Dirichlet-energy regularized LoRAModel=Qwen2.5-VL-7B2026.05 | 48.3 | — | |
| Naive LoRA fine-tuningModel=InternVL3-8B2026.05 | 45.2 | — | |
| 2D regularizationModel=InternVL3-8B2026.05 | 44.7 | — | |
| Text-domain cog mapModel=InternVL3-8B2026.05 | 43.1 | — | |
| Naive LoRA fine-tuningModel=Qwen2.5-VL-7B2026.05 | 41.3 | — | |
| 2D regularizationModel=Qwen2.5-VL-7B2026.05 | 39.8 | — | |
| Text-domain cog mapModel=Qwen2.5-VL-7B2026.05 | 38.5 | — | |
| MiMo-EmbodiedNumber of Parameters=7B2026.04 | 36.2 | — | |
| QwenModel Version=3-VL, Number of Parameters=4B2026.04 | 31 | — | |
| QwenModel Version=3-VL, Number of Parameters=2B2026.04 | 28.4 | — | |
| RoboBrainModel Version=2.5, Number of Parameters=4B2026.04 | 26.9 | — | |
| GeminiModel Version/Scale=3.0 Pro2026.04 | — | 66 | |
| HY-EmbodiedModel Version/Scale=0.5 MoE A32B2026.04 | — | 69.2 | |
| KimiModel Version/Scale=K2.52026.04 | — | 57.8 | |
| QwenModel Version/Scale=3.5 A17B2026.04 | — | 59 | |
| SeedModel Version/Scale=2.02026.04 | — | 55.2 |