3D Visual Grounding on Nr3D without GT object class
72.5Easy SuccessSceneVerse
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SceneVerseLearning Setting=Fully-Supervised, LLM/VLM=-, w/o PC.=✗2026.04 | 72.5 | 57.8 | 56.9 | 67.9 | 64.9 | |
| TABLearning Setting=Zero-Shot, LLM/VLM=Qwen3-VL-32B, w/o PC.=✓2026.04 | 72.1 | 63.2 | 62.5 | 71.4 | 68 | |
| TABLearning Setting=Zero-Shot, LLM/VLM=Qwen3-VL-32B, w/o PC.=✗2026.04 | 72.1 | 61.4 | 63.5 | 69.5 | 67.2 | |
| SPAZERLearning Setting=Zero-Shot, LLM/VLM=GPT-4o, w/o PC.=✗2026.04 | 68 | 58.8 | 59.9 | 66.2 | 63.8 | |
| EDALearning Setting=Fully-Supervised, LLM/VLM=-, w/o PC.=✗2026.04 | 58.2 | 46.1 | 50.2 | 53.1 | 52.1 | |
| SeqVLMLearning Setting=Zero-Shot, LLM/VLM=Doubao-1.5-pro, w/o PC.=✗2026.04 | 58.1 | 47.4 | 51 | 54.5 | 53.2 | |
| VLM-GrounderLearning Setting=Zero-Shot, LLM/VLM=GPT-4o, w/o PC.=✓2026.04 | 55.2 | 39.5 | 45.8 | 49.4 | 48 | |
| BUTD-DETRLearning Setting=Fully-Supervised, LLM/VLM=-, w/o PC.=✗2026.04 | 54.6 | 60.7 | 48.4 | 46 | 58 | |
| SeeGroundLearning Setting=Zero-Shot, LLM/VLM=Qwen2-VL-72B, w/o PC.=✗2026.04 | 54.5 | 38.3 | 42.3 | 48.2 | 46.1 | |
| 3DVG-TLearning Setting=Fully-Supervised, LLM/VLM=-, w/o PC.=✗2026.04 | 48.5 | 34.8 | 34.8 | 43.7 | 40.8 | |
| ZSVG3DLearning Setting=Zero-Shot, LLM/VLM=GPT-4 turbo, w/o PC.=✗2026.04 | 46.5 | 31.7 | 36.8 | 40 | 39 | |
| InstanceReferLearning Setting=Fully-Supervised, LLM/VLM=-, w/o PC.=✗2026.04 | 46 | 31.8 | 34.5 | 41.9 | 38.8 | |
| ReferIt3DNetLearning Setting=Fully-Supervised, LLM/VLM=-, w/o PC.=✗2026.04 | 43.6 | 27.9 | 32.5 | 37.1 | 35.6 |