Research map / Vision-language and multimodal / Multimodal LLMs
Where to Look Matters: Learning Influential Views for VLM-based 3D Visual Grounding
Tsung-Chih Chiang, Hsuan-Kung Yang, Jou-Min Liu, Ting-Ru Liu, Chun-Wei Huang, Quan Kong, Chun-Yi Lee
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 |
|---|---|---|---|
| use visibility-based view selection instead of random frames (ablation) | IVSGround pipeline · NR3D 250-query subset | overall accuracy: 58.4 -> 64.8 | Improved |
| add learned influential view selection (IVS) on top of visibility selection (ablation) | IVSGround pipeline · NR3D 250-query subset | overall accuracy: 64.8 -> 66.0 | Improved |
| mix 2 IVS views with 3 visibility views rather than pure heuristic (ablation) | IVSGround (k=5) · NR3D 250-query subset | overall accuracy: 64.8 -> 68.8 | Improved |
| use pure IVS (5 views) instead of the hybrid mix (ablation) | IVSGround (k=5) · NR3D 250-query subset | overall accuracy: 68.8 -> 63.6 | Hurt |
| add VLM-based verification stage to Cascade Object Screening (ablation) | IVSGround COS · NR3D 250-query subset | overall accuracy: 55.2 -> 58.4 | Improved |
| add VLM-based verification stage to Cascade Object Screening (ablation) | IVSGround COS · NR3D 250-query subset (Hard) | Hard accuracy: 43.9 -> 48.2 | Improved |
| plug IVS views into SeqVLM without retraining (ablation) | SeqVLM · NR3D 250-query subset | Acc@0.25: 52.0 -> 58.4 (+6.4) | Improved |
| add IVS to Qwen3-VL-32B reasoner (ablation) | IVSGround · NR3D 250-query subset | overall accuracy: 64.8 -> 66.0 | Improved |
| add IVS (selector trained on Qwen3-VL-32B feedback) to Qwen2.5-VL-72B reasoner without retraining (ablation) | IVSGround · NR3D 250-query subset | overall accuracy: 55.2 -> 56.8 | Improved |
| IVSGround vs SPAZER under shared Qwen2.5-VL-72B reasoner | IVSGround · NR3D 250-query subset | Acc@0.25: 57.6 vs 56.0 | Improved |
| IVSGround vs SeeGround under shared Qwen2.5-VL-72B reasoner | IVSGround · NR3D 250-query subset | Acc@0.25: 57.6 vs 54.0 | Improved |
| IVSGround vs SPAZER on ScanRefer VLM-Grounder subset | IVSGround · ScanRefer (VLM-Grounder subset) | Acc@0.25 overall: 58.0 vs 57.2 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SPAZER — SPAZER: Spatial-Semantic Progressive Reasoning Agent for Zero-shot 3D Visual Grounding (NeurIPS 2025) · better on 5 of 5 shared results
- SeeGround — SeeGround: See and Ground for Zero-Shot Open-Vocabulary 3D Visual Grounding (CVPR 2025) · better on 3 of 3 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
- SPAZER: Spatial-Semantic Progressive Reasoning Agent for Zero-shot 3D Visual Grounding (NeurIPS 2025 · similarity 0.92)
- From Objects to Anywhere: A Holistic Benchmark for Multi-level Visual Grounding in 3D Scenes (NeurIPS 2025 · similarity 0.90)
- Grounded 3D-Aware Spatial Vision-Language Modeling (CVPR 2026 · similarity 0.90)
- SeeGround: See and Ground for Zero-Shot Open-Vocabulary 3D Visual Grounding (CVPR 2025 · similarity 0.90)
- MonoVLM: Monocular 3D Visual Grounding with Vision Language Models (CVPR 2026 · similarity 0.90)
- ReasonGrounder: LVLM-Guided Hierarchical Feature Splatting for Open-Vocabulary 3D Visual Grounding and Reasoning (CVPR 2025 · similarity 0.90)
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.
