Research map / Point clouds / Point cloud classification
Direction-aware 3D Large Multimodal Models
Quan Liu, Weihao Xuan, Junjue Wang, Naoto Yokoya, Ling Shao, Shijian Lu
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 PoseAlign-T (ego-aligned point cloud) | 3D-LLAVA · ScanRefer val | mIoU: 42.6 -> 55.4 | Improved |
| add PoseAlign-T | 3D-LLAVA · Multi3DRefer val | mIoU: 48.1 -> 54.3 | Improved |
| add PoseAlign-T | 3D-LLAVA · Scan2Cap val | LLM-as-judge accuracy: 28.1 -> 31.4 | Improved |
| add PoseAlign-T | 3D-LLAVA · Scan2Cap val | CIDEr@0.5: 77.4 -> 76.1 | Hurt |
| add PoseAlign-E | Chat-Scene · ScanQA val | LLM-as-judge accuracy: 43.9 -> 44.7 | Improved |
| add PoseAlign-T | LL3DA · ScanQA val | LLM-as-judge accuracy: 34.7 -> 35.6 | Improved |
| use PoseRecover poses rather than random poses (ablation) | 3D-LLAVA + PoseAlign · ScanRefer val | mIoU: 39.0 -> 55.4 | Improved |
| use transform rather than embedding the pose (PoseAlign-E) (ablation) | 3D-LLAVA · ScanRefer val | mIoU: 43.2 -> 68.5 | Improved |
| use transform rather than pose text prompt (PoseAlign-P) (ablation) | 3D-LLAVA · ScanRefer val | mIoU: 44.2 -> 68.5 | Improved |
| add Clip ratio X=0.3 instead of Top pose selection (ablation) | 3D-LLAVA + PoseAlign-T · Scan2Cap val | LLM-as-judge accuracy: 29.7 -> 31.4 | Mixed |
| vary Clip ratio X from 0.0 to 0.49 (ablation) | 3D-LLAVA + PoseAlign-T · ScanQA val | LLM-as-judge accuracy: 47.3 -> 46.9 | No clear effect |
| clip ratio X=0.4 or higher (Chat-Scene PoseAlign-E) (ablation) | Chat-Scene · Scan2Cap val | LLM-as-judge accuracy: 26.9 -> 6.2 | Hurt |
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.
- LL3DA — LL3DA: Visual Interactive Instruction Tuning for Omni-3D Understanding Reasoning and Planning (CVPR 2024) · better on 1 of 1 shared result
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
- Vid-LLM: A Compact Video-based 3D Multimodal LLM with Reconstruction–Reasoning Synergy (ICLR 2026 · similarity 0.87)
- SpatialLLM: A Compound 3D-Informed Design towards Spatially-Intelligent Large Multimodal Models (CVPR 2025 · similarity 0.87)
- Towards Comprehensive Scene Understanding: Integrating First and Third-Person Views for LVLMs (NeurIPS 2025 · similarity 0.87)
- Video-3D LLM: Learning Position-Aware Video Representation for 3D Scene Understanding (CVPR 2025 · similarity 0.87)
- LL3DA: Visual Interactive Instruction Tuning for Omni-3D Understanding Reasoning and Planning (CVPR 2024 · similarity 0.86)
- Spatial Reasoning with Vision-Language Models in Ego-Centric Multi-View Scenes (ICLR 2026 · similarity 0.86)
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.
