Embodied Referring on RefSpatial-Bench average
44.22AccuracyGeoThinker
Evaluation Results
| Method | Links | |
|---|---|---|
| GeoThinkerSFT Status=2B-SFT2026.02 | 44.22 | |
| RoboReferSFT Status=2B-SFT2026.02 | 42.56 | |
| Gemini-2.5-ProModel Category=Proprietary Models2026.02 | 32.77 | |
| Molmo-72BModel Category=Referring Specialist Models2026.02 | 27.25 | |
| Molmo-7BModel Category=Referring Specialist Models2026.02 | 15.66 | |
| RoboPointModel Category=Referring Specialist Models2026.02 | 13.51 | |
| SpaceLLaVAModel Category=Referring Specialist Models2026.02 | 4.71 |