Research map / 3D vision / 3D generation
3DGraphLLM: Combining Semantic Graphs and Large Language Models for 3D Scene Understanding
Tatiana Zemskova, Dmitry Yudin
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 2 semantic-relation edges per object (ablation) | 3DGraphLLM LLAMA3-8B · ScanRefer (Mask3D) | Acc@0.5: 52.0 -> 54.3 | Improved |
| add 2 semantic-relation edges per object (ablation) | 3DGraphLLM LLAMA3-8B · Multi3DRefer (Mask3D) | F1@0.5: 55.1 -> 57.3 | Improved |
| add 2 semantic-relation edges per object (ablation) | 3DGraphLLM Vicuna-1.5-7B · ScanRefer (Mask3D) | Acc@0.5: 50.2 -> 50.1 | No clear effect |
| add 2 semantic-relation edges per object (ablation) | 3DGraphLLM Vicuna-1.5-7B · ScanQA | CIDEr: 87.7 -> 92.2 | Improved |
| add 2 semantic-relation edges per object (ablation) | 3DGraphLLM · ScanRefer (GT segmentation) | Acc@0.5: 61.5 -> 66.9 | Improved |
| add 2 semantic-relation edges per object (ablation) | 3DGraphLLM · ScanRefer (OneFormer3D) | Acc@0.5: 50.0 -> 52.8 | Improved |
| add pre-training on GT instance segmentation (ablation) | 3DGraphLLM LLAMA3-8B · ScanRefer | Acc@0.5: 54.3 -> 56.2 | Improved |
| add 3RScan data to pre-training (ablation) | 3DGraphLLM LLAMA3-8B · ScanRefer | Acc@0.5: 56.2 -> 56.6 | Improved |
| add NMS filter on neighbours (ablation) | 3DGraphLLM LLAMA3-8B · ScanRefer (Mask3D) | Acc@0.5: 55.6 -> 55.7 | Improved |
| add 1 cm minimum neighbour distance (ablation) | 3DGraphLLM LLAMA3-8B · ScanRefer (Mask3D) | Acc@0.5: 55.7 -> 56.2 | Improved |
| represent relations as triplets (ablation) | 3DGraphLLM LLAMA3-8B · Multi3DRefer (Mask3D) | F1@0.5: 56.3 -> 57.3 | Improved |
| add semantic edges (ablation) | 3DGraphLLM LLAMA3-8B · Scan2Cap | CIDEr@0.5: 80.0 -> 85.6 | Mixed |
Later papers that compared with this method
1 later accepted paper reports 3DGraphLLM as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Scalable Object Relation Encoding for Better 3D Spatial Reasoning in Large Language Models (CVPR 2026 · 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
- Open-Vocabulary Functional 3D Scene Graphs for Real-World Indoor Spaces (CVPR 2025 · similarity 0.91)
- Object-Centric Representation Learning for Enhanced 3D Semantic Scene Graph Prediction (NeurIPS 2025 · similarity 0.90)
- Open3DSG: Open-Vocabulary 3D Scene Graphs from Point Clouds with Queryable Objects and Open-Set Relationships (CVPR 2024 · similarity 0.90)
- 3D-PLOT-LLM: Part-Level Object Tokens for 3D Large Language Models (NeurIPS 2026 · similarity 0.90)
- LSceneLLM: Enhancing Large 3D Scene Understanding Using Adaptive Visual Preferences (CVPR 2025 · similarity 0.90)
- Chat-Scene: Bridging 3D Scene and Large Language Models with Object Identifiers (NeurIPS 2024 · 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.
