Research map / 3D vision / 3D generation
Think While You Map: Asynchronous Vision-Language Agents for Incremental 3D Scene Graphs
Deniz Bickici, Michael Pabst, Shohei Mori, Dieter Schmalstieg
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 |
|---|---|---|---|
| replace RAM++ with Qwen3-VL as frontend tagger (ablation) | ThinkGraphs backend · Replica | mAcc: 0.37 -> 0.46 | Improved |
| add text-guided embedding selection (ablation) | ThinkGraphs backend · Replica | mIoU: 0.26 -> 0.35 | Improved |
| add adaptive area filter (ablation) | ThinkGraphs backend · Replica | mAcc: 0.52 -> 0.58 | Improved |
| replace SBERT with CLIP text-label embeddings (ablation) | ThinkGraphs backend · Replica | mAcc: 0.46 -> 0.47 | Improved |
| add Critic Agent (ablation) | ThinkGraphs base · Nr3D | A@0.25: 38.5 -> 39.3 | Improved |
| add Description Agent (ablation) | ThinkGraphs base · Nr3D (hard) | A@0.25: +3.5 A@0.25 | Improved |
| add both Critic and Description agents (ablation) | ThinkGraphs base · Nr3D | A@0.25: 38.5 -> 41.8 | Improved |
| add both agents (ablation) | ThinkGraphs base · Nr3D (hard) | A@0.25: 18.8 -> 25.9 | Improved |
| use Qwen3-VL tagger instead of RAM++ in full system (ablation) | ThinkGraphs · Nr3D | A@0.25: 33.6 -> 41.8 | Improved |
| use GPT-5.4 instead of GPT-4o as evaluator | ThinkGraphs · Nr3D | A@0.25: 41.8 -> 45.4 | Improved |
| use Llama-4 Maverick instead of GPT-4o as evaluator | ThinkGraphs · Nr3D | A@0.25: 41.8 -> 39.9 | Hurt |
| increase map coverage from 10% to 100% | ThinkGraphs · ScanNet | mIoU: 0.09 -> 0.44 | Improved |
Closest papers by meaning
- View-on-Graph: Zero-Shot 3D Visual Grounding via Vision-Language Reasoning on Scene Graphs (AAAI 2026 · similarity 0.90)
- Beyond Isolated Objects: Relationship-aware Open Vocabulary Scene Understanding via 3D Scene Graph Analysis (ECCV 2026 · similarity 0.90)
- GraphWrit3R: End-to-End Writing of Scene Graphs for Multi-Modal 3D Scenes (NeurIPS 2026 · similarity 0.90)
- Open-World 3D Scene Graph Generation for Retrieval-Augmented Reasoning (AAAI 2026 · similarity 0.89)
- Uncovering and Shaping the Latent Representation of 3D Scene Topology in Vision-Language Models (NeurIPS 2026 · similarity 0.89)
- IR3D-Bench: Evaluating Vision-Language Model Scene Understanding as Agentic Inverse Rendering (NeurIPS 2025 · similarity 0.88)
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.
