3D Question Answering on SQA3D (val)
60.8EMLEO-VL
Evaluation results
| Method | Links | |
|---|---|---|
| LEO-VLVision modal.=D,I, # of tokens=7502026.05 | 60.8 | |
| 3DRSVision modal.=I, # of tokens=80002026.05 | 60.6 | |
| Video-3D-LLMVision modal.=I, # of tokens=80002026.05 | 58.6 | |
| Proxy3DVision modal.=I, # of tokens=700, Backbone=Qwen2.5-VL-7B2026.05 | 57.5 | |
| GPT4Scene (HDM)Vision modal.=B,I, # of tokens=80002026.05 | 57.4 | |
| Spatial-MLLM-4BVision modal.=I, # of tokens=31002026.05 | 55.9 | |
| Descrip3DVision modal.=I2026.05 | 55.7 | |
| LLaVA-3DVision modal.=D,I, # of tokens=30962026.05 | 55.6 | |
| Chat-SceneVision modal.=P2026.05 | 54.6 | |
| LEOVision modal.=P,I2026.05 | 50 | |
| ScanQAVision modal.=P, # of tokens=2562026.05 | 47.2 | |
| Qwen2-VL-7BVision modal.=I, # of tokens=80002026.05 | 40.7 | |
| SpaceEra++Backbone=Qwen2.5-VL-7B2026.07 | 30.5 | |
| +SpaceAlignBackbone=Qwen2.5-VL-7B2026.07 | 28.5 | |
| +ScenePickBackbone=Qwen2.5-VL-7B2026.07 | 26.3 | |
| SpaceEraBackbone=Qwen2.5-VL-7B2026.07 | 24.5 | |
| Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2026.07 | 17.2 |