Research map / Vision-language and multimodal / CLIP and zero-shot
View-on-Graph: Zero-Shot 3D Visual Grounding via Vision-Language Reasoning on Scene Graphs
Yuanyuan Liu, Haiyang Mei, Dongyang Zhan, Jiayue Zhao, Dongsheng Zhou, Bo Dong, Xin Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add graph traversal (candidate filtering) (ablation) | VoG · ScanRefer val | Acc@0.25: 30.3 -> 35.9 | Improved |
| add context integration (view stitching + object pool) (ablation) | VoG · ScanRefer val | Acc@0.25: 32.7 -> 35.9 | Improved |
| add multi-round reasoning on top of full structure (ablation) | VoG · ScanRefer val | Acc@0.25: 32.1 -> 35.9 | Improved |
| multi-round random view selection without traversal or pool vs one round (ablation) | VoG · ScanRefer val | Acc@0.25: 32.1 -> 32.6 | Improved |
| multi-round reasoning on minimal structure (S4) without pool vs one round (ablation) | VoG · ScanRefer val | Acc@0.25: 31.6 -> 31.7 | No clear effect |
| multi-round with pool on minimal structure S4 (ablation) | VoG · ScanRefer val | Acc@0.25: 31.6 -> 27.9 | Hurt |
| remove all structure (S7) vs full structure (ablation) | VoG · ScanRefer val | Acc@0.25: 32.1 -> 30.2 | Hurt |
| increase max reasoning depth beyond 4 (ablation) | VoG · ScanRefer val | Acc@0.25 | No clear effect |
| increase view nodes beyond 32 (ablation) | VoG · ScanRefer val | Acc@0.25 | Hurt |
| let VLM directly choose initial view instead of fuzzy matching (ablation) | VoG · ScanRefer val | Acc@0.25 | Hurt |
| scale VLM from Qwen2-VL-2B to 72B | VoG · ScanRefer val | Acc@0.25: 35.9 -> 44.8 | Improved |
| VoG vs SeeGround at 2B | SeeGround · ScanRefer val | Acc@0.25: 30.2 -> 35.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SeeGround — SeeGround: See and Ground for Zero-Shot Open-Vocabulary 3D Visual Grounding (CVPR 2025) · better on 11 of 12 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Chain of Semantics Programming in 3D Gaussian Splatting Representation for 3D Vision Grounding (CVPR 2025 · similarity 0.92)
- SeeGround: See and Ground for Zero-Shot Open-Vocabulary 3D Visual Grounding (CVPR 2025 · similarity 0.92)
- SPAZER: Spatial-Semantic Progressive Reasoning Agent for Zero-shot 3D Visual Grounding (NeurIPS 2025 · similarity 0.91)
- Visual Programming for Zero-shot Open-Vocabulary 3D Visual Grounding (CVPR 2024 · similarity 0.91)
- S$^2$-MLLM: Boosting Spatial Reasoning Capability of MLLMs for 3D Visual Grounding with Structural Guidance (CVPR 2026 · similarity 0.91)
- Grounded 3D-Aware Spatial Vision-Language Modeling (CVPR 2026 · similarity 0.91)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
