3D Visual Grounding on ScanRefer 10 (val)
62.9Accuracy @ IoU=0.25 (Overall)3DRS
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| 3DRSSupervision=Fully Supervised2026.06 | 62.9 | 87.4 | 77.9 | 57 | 50.8 | 56.1 | |
| Ross3DSupervision=Fully Supervised2026.06 | 61.1 | 87.2 | 77.4 | 54.8 | 48.9 | 54.4 | |
| Video-3D LLMSupervision=Fully Supervised2026.06 | 58.1 | 88 | 78.3 | 50.9 | 45.3 | 51.7 | |
| Agentic Collaborative CognitionSupervision=Zero-shot (full)2026.06 | 58.1 | 81.2 | 72.9 | 50.8 | 43.9 | 50.9 | |
| Agentic Collaborative CognitionSupervision=Zero-shot (250 queries)2026.06 | 58 | 83 | 74.5 | 52.2 | 44.3 | 50 | |
| SPAZERSupervision=Zero-shot (250 queries)2026.06 | 57.2 | 80.9 | 72.3 | 51.7 | 43.4 | 48.8 | |
| SeqVLMSupervision=Zero-shot (250 queries)2026.06 | 55.6 | 77.3 | 72.7 | 47.8 | 41.3 | 49.6 | |
| BUTD-DETRSupervision=Fully Supervised2026.06 | 52.2 | 84.2 | 66.3 | 46.6 | 35.1 | 39.8 | |
| VLM-GrounderSupervision=Zero-shot (250 queries)2026.06 | 51.6 | 66 | 29.8 | 48.3 | 33.5 | 32.8 | |
| 3D-VisTASupervision=Fully Supervised2026.06 | 50.6 | 81.6 | 75.1 | 43.7 | 39.1 | 45.8 | |
| CSVGSupervision=Zero-shot (full)2026.06 | 49.6 | 68.8 | 61.2 | 38.4 | 27.3 | 39.8 | |
| 3DVG-TSupervision=Fully Supervised2026.06 | 45.9 | 77.2 | 58.5 | 38.4 | 28.7 | 34.5 | |
| SeeGroundSupervision=Zero-shot (full)2026.06 | 44.1 | 75.7 | 68.9 | 34 | 30 | 39.4 | |
| ScanReferSupervision=Fully Supervised2026.06 | 39 | 67.6 | 46.2 | 32.1 | 21.3 | 26.1 | |
| ZSVG3DSupervision=Zero-shot (full)2026.06 | 36.4 | 63.8 | 58.4 | 27.7 | 24.6 | 32.7 | |
| LLM-GrounderSupervision=Zero-shot (full)2026.06 | 17.1 | — | — | — | — | 5.3 |