Research map / 3D vision / 3D generation
Scalable Object Relation Encoding for Better 3D Spatial Reasoning in Large Language Models
Shengli Zhou, Minghang Zheng, Feng Zheng, Yang Liu
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 QuatRoPE via IGRE | Chat-Scene · ScanRefer (GT segmentation) | Acc@0.5: 50.3 -> 55.0 | Improved |
| add QuatRoPE via IGRE | 3DGraphLLM · ScanRefer (GT segmentation) | Acc@0.5: 55.8 -> 58.2 | Improved |
| add QuatRoPE | 3DGraphLLM-7B · ScanRefer (Mask3D) | Acc@0.5: 51.3 -> 52.5 | Improved |
| add QuatRoPE | Chat-Scene · ASR benchmark | Acc@0.25: 22.92 -> 27.38 | Improved |
| add QuatRoPE | 3DGraphLLM (Llama-3-8B) · ASR benchmark | Acc@0.25: 37.50 -> 41.96 | Improved |
| use IGRE instead of Trans-Additive composition (ablation) | QuatRoPE + Chat-Scene · ScanRefer Acc@0.5 (Llama-3.2-1B) | Acc@0.5: 52.79 -> 55.00 | Improved |
| use Trans-Additive composition for QuatRoPE (ablation) | 3DGraphLLM · ScanRefer Acc@0.5 (Llama-3.2-1B) | Acc@0.5: 55.75 -> 53.38 | Hurt |
| use QuatRoPE instead of M-RoPE (ablation) | 3DGraphLLM · ScanRefer Acc@0.5 (Llama-3.2-1B) | Acc@0.5: 57.48 -> 58.15 | Improved |
| add raw coordinates to features (ablation) | 3DGraphLLM · ScanRefer Acc@0.5 (Llama-3.2-1B) | Acc@0.5: 55.75 -> 3.44 | Hurt |
| add raw coordinates to features (ablation) | Chat-Scene · ScanRefer Acc@0.5 (Llama-3.2-1B) | Acc@0.5: 50.33 -> 52.01 | Improved |
| holistic quaternion encoding on severe false-nearby cases (ablation) | 3DGraphLLM-1B · ScanRefer (aspect ratio < 0.05) | accuracy: 84.62 -> 92.31 | 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.
- 3DGraphLLM — 3DGraphLLM: Combining Semantic Graphs and Large Language Models for 3D Scene Understanding (ICCV 2025) · better on 6 of 6 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
- SpatialRGPT: Grounded Spatial Reasoning in Vision-Language Models (NeurIPS 2024 · similarity 0.90)
- Spatial Understanding from Videos: Structured Prompts Meet Simulation Data (NeurIPS 2025 · similarity 0.90)
- Pursuing Minimal Sufficiency in Spatial Reasoning (ICLR 2026 · similarity 0.89)
- SpaceMind: Camera-Guided Modality Fusion for Spatial Reasoning in Vision-Language Models (CVPR 2026 · similarity 0.89)
- SpatialReasoner: Towards Explicit and Generalizable 3D Spatial Reasoning (NeurIPS 2025 · similarity 0.89)
- From Flatland to Space: Teaching Vision-Language Models to Perceive and Reason in 3D (NeurIPS 2025 · similarity 0.89)
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.
