3D Visual Grounding on ScanRefer Unique
90.3Acc@0.25 (IoU=0.25)GPT4Scene
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT4SceneSupervision=Supervised, VLM=None2025.12 | 90.3 | 83.7 | |
| GPT4SceneLLM/VLM=Qwen2-VL-7B, w/o PC=✗2026.04 | 90.3 | 83.7 | |
| TABLLM/VLM=Qwen3-VL-32B, w/o PC=✓2026.04 | 90.2 | 57.6 | |
| TABLLM/VLM=Qwen3-VL-32B, w/o PC=✗2026.04 | 90.2 | 77.2 | |
| Chat-SceneLLM/VLM=Vicuna-7B, w/o PC=✗2026.04 | 89.5 | 82.4 | |
| G3-LQSupervision Level=Fully-Supervised, Venue=CVPR’24, Agent=—2025.12 | 88.6 | 73.3 | |
| G3-LQLLM/VLM=-, w/o PC=✗2026.04 | 88.6 | 73.3 | |
| Video-3D-LLMLLM/VLM=LLaVA-Video-7B, w/o PC=✗2026.04 | 87.9 | 78.3 | |
| 3D-RSLLM/VLM=LLaVA-Next-Video, w/o PC=✗2026.04 | 87.4 | 77.9 | |
| MCLNSupervision Level=Fully-Supervised, Venue=ECCV’24, Agent=—2025.12 | 86.9 | 72.7 | |
| ConcreteNetSupervision Level=Fully-Supervised, Venue=ECCV’24, Agent=—2025.12 | 86.4 | 82.1 | |
| EDASupervision Level=Fully-Supervised, Venue=CVPR’23, Agent=—2025.12 | 85.8 | 68.6 | |
| EDALLM/VLM=-, w/o PC=✗2026.04 | 85.8 | 68.6 | |
| BUTD-DETRSupervision Level=Fully-Supervised, Venue=ECCV’22, Agent=—2025.12 | 84.2 | 66.3 | |
| BUTD-DETRLLM/VLM=-, w/o PC=✗2026.04 | 84.2 | 66.3 | |
| ViewSRDSupervision=Supervised, VLM=None2025.12 | 82.1 | 68.2 | |
| 3D-VisTASupervision Level=Fully-Supervised, Venue=ICCV’23, Agent=—2025.12 | 81.6 | 75.1 | |
| SPAZERSupervision=Zero-Shot, VLM=GPT-4o2025.12 | 80.9 | 72.3 | |
| SPAZERLLM/VLM=GPT-4o, w/o PC=✗2026.04 | 80.9 | 72.3 | |
| VoGSupervision Level=Zero-Shot, Venue=Ours, Agent=Qwen2-VL-72B2025.12 | 78.6 | 71.5 | |
| OpenGroundSupervision=Zero-Shot, VLM=GLM-4.5V2025.12 | 77.8 | 74.4 | |
| InstanceReferSupervision Level=Fully-Supervised, Venue=ICCV’21, Agent=—2025.12 | 77.5 | 66.8 | |
| SeqVLMSupervision=Zero-Shot, VLM=Doubao-1.5-pro2025.12 | 77.3 | 72.7 | |
| SeqVLMLLM/VLM=Doubao-1.5-pro, w/o PC=✗2026.04 | 77.3 | 72.7 | |
| 3DVG-TSupervision Level=Fully-Supervised, Venue=ICCV’21, Agent=—2025.12 | 77.2 | 58.5 | |
| 3DVG-TLLM/VLM=-, w/o PC=✗2026.04 | 77.2 | 58.5 | |
| SeeGroundSupervision=Zero-Shot, VLM=Qwen2-VL-72b2025.12 | 75.7 | 68.9 | |
| SeeGroundSupervision Level=Zero-Shot, Venue=CVPR’25, Agent=Qwen2-VL-72B2025.12 | 75.7 | 68.9 | |
| SeeGroundLLM/VLM=Qwen2-VL-72B, w/o PC=✗2026.04 | 75.7 | 68.9 | |
| VoGSupervision Level=Zero-Shot, Venue=Ours, Agent=Qwen2-VL-2B2025.12 | 69.1 | 63.1 | |
| CSVGLLM/VLM=Mistral-Large-2407, w/o PC=✗2026.04 | 68.8 | 61.2 | |
| ScanReferSupervision=Supervised, VLM=None2025.12 | 67.6 | 46.2 | |
| ScanReferSupervision Level=Fully-Supervised, Venue=ECCV’20, Agent=—2025.12 | 67.6 | 46.2 | |
| ScanReferLLM/VLM=-, w/o PC=✗2026.04 | 67.6 | 46.2 | |
| VLM-GrounderLLM/VLM=GPT-4o, w/o PC=✓2026.04 | 66 | 29.8 | |
| ZSVG3DSupervision=Zero-Shot, VLM=GPT-4 turbo2025.12 | 63.8 | 58.4 | |
| ZSVG3DSupervision Level=Zero-Shot, Venue=CVPR’24, Agent=GPT-4 Turbo2025.12 | 63.8 | 58.4 | |
| ZSVG3DLLM/VLM=GPT-4 turbo, w/o PC=✗2026.04 | 63.8 | 58.4 | |
| SeeGroundSupervision Level=Zero-Shot, Venue=CVPR’25, Agent=Qwen2-VL-2B2025.12 | 59.9 | 55 | |
| VLM-GrounderSupervision=Zero-Shot, VLM=GPT-4o, notes=results on selected 250 samples2025.12 | 51.6 | 32.8 | |
| OpenSceneSupervision Level=Zero-Shot, Venue=CVPR’23, Agent=CLIP2025.12 | 20.1 | 13.1 |