Research map / 3D vision / 3D generation
Chat-Scene++: Exploiting Context-Rich Object Identification for 3D LLM
Haifeng Huang, Yilun Chen, Zheng Wang, Jiangmiao Pang, Zhou Zhao
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 learnable object identifier tokens (ablation) | Chat-Scene++ · ScanRefer | Acc@0.5: 25.5 -> 52.6 | Improved |
| use learnable ID tokens instead of fixed Gaussian embeddings (ablation) | Chat-Scene++ · ScanRefer | Acc@0.5: 48.3 -> 52.6 | Improved |
| use learnable single-token IDs instead of plain text IDs (ablation) | Chat-Scene++ · ScanRefer | Acc@0.5: 49.2 -> 52.6 | Improved |
| randomize object ID order during training (ablation) | Chat-Scene++ · ScanRefer | Acc@0.5: 52.6 -> 53.4 | Improved |
| enhance Mask3D features with CLASP pre-training (ablation) | Chat-Scene++ · ScanRefer | Acc@0.5: 37.6 -> 47.3 | Improved |
| add position embeddings to Uni3D features (ablation) | Chat-Scene++ · ScanRefer | Acc@0.5: 41.2 -> 41.5 | No clear effect |
| add DINOv2 2D features to CLASP 3D features (ablation) | Chat-Scene++ · ScanRefer | Acc@0.5: 47.3 -> 53.4 | Improved |
| use DINOv2 instead of CLIP as 2D encoder (ablation) | Chat-Scene++ · ScanRefer | Acc@0.5: 49.2 -> 53.4 | Improved |
| add QA CoT data (ablation) | Chat-Scene++ · ScanQA | CIDEr: 89.3 -> 93.1 | Improved |
| add category-level grounded CoT data (ablation) | Chat-Scene++ · ScanRefer | Acc@0.5: 52.1 -> 52.9 | Improved |
| add space-level grounded CoT data (ablation) | Chat-Scene++ · ScanRefer | Acc@0.5: 52.1 -> 51.8 | No clear effect |
| add extra second training stage with CoT data (ablation) | Chat-Scene++ · ScanQA / ScanRefer | CIDEr / Acc@0.5: 93.5 -> 92.9 CIDEr; 53.4 -> 53.8 Acc@0.5 | Mixed |
Closest papers by meaning
- Chat-Scene: Bridging 3D Scene and Large Language Models with Object Identifiers (NeurIPS 2024 · similarity 0.94)
- MMScan: A Multi-Modal 3D Scene Dataset with Hierarchical Grounded Language Annotations (NeurIPS 2024 · similarity 0.90)
- LSceneLLM: Enhancing Large 3D Scene Understanding Using Adaptive Visual Preferences (CVPR 2025 · similarity 0.90)
- SceneScaffold: Active Scene-State Construction for Unified 3D Scene Understanding (NeurIPS 2026 · similarity 0.90)
- Video-3D LLM: Learning Position-Aware Video Representation for 3D Scene Understanding (CVPR 2025 · similarity 0.90)
- Language-Image Models with 3D Understanding (ICLR 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.
