Geometric Understanding on ScanNet (test)
97.5RR@0.05mDINOv3 + SpatialBoost
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DINOv3 + SpatialBoostSpatialBoost=true2026.03 | 97.5 | 0.06 | |
| V-JEPAv2Encoder Type=Vision-only2026.03 | 95.4 | 0.08 | |
| DINOv2 + SpatialBoostSpatialBoost=true2026.03 | 92.4 | 0.13 | |
| DINOv32026.03 | 86.9 | 0.1 | |
| SigLIPv2 + SpatialBoostSpatialBoost=true2026.03 | 86.4 | 0.15 | |
| TIPSEncoder Type=Vision-Language2026.03 | 84 | 0.16 | |
| DINOv22026.03 | 82.4 | 0.15 | |
| dino.txtEncoder Type=Vision-Language2026.03 | 81.7 | 0.16 | |
| OpenCLIP + SpatialBoostSpatialBoost=true2026.03 | 78.8 | 0.17 | |
| AIMv2Encoder Type=Vision-Language2026.03 | 47.8 | 0.31 | |
| SigLIPv22026.03 | 47.8 | 0.28 | |
| PE-CoreEncoder Type=Vision-Language2026.03 | 44.6 | 0.29 | |
| OpenCLIP2026.03 | 22.6 | 0.4 |