Research map / Vision-language and multimodal / CLIP and zero-shot
Agent3D-Zero: An Agent for Zero-shot 3D Understanding
sha zhang, Di Huang, Jiajun Deng, Shixiang Tang, Wanli Ouyang, Tong He, Yanyong Zhang
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 |
|---|---|---|---|
| select viewpoints iteratively with the VLM instead of randomly (ablation) | Agent3D-Zero · ScanQA val | CIDEr: 40.0 -> 71.8 | Improved |
| select viewpoints iteratively instead of randomly (ablation) | Agent3D-Zero · ScanNet v2 semantic segmentation | mIoU: 5.6 -> 8.7 | Improved |
| increase number of observation viewpoints from 6 to 24 (ablation) | Agent3D-Zero · ScanQA val (20% subset) | CIDEr: 50.8 -> 82.0 | Improved |
| increase Set-of-Line grid density from 4x4 to 8x8 (ablation) | Agent3D-Zero · ScanQA val (20% subset) | CIDEr: 66.0 -> 82.0 | Improved |
| remove grid lines (0x0) or use too dense lines (16x16) (ablation) | Agent3D-Zero · ScanQA val (20% subset) | all metrics: GPT-4V fails to output results | Hurt |
| zero-shot Agent3D-Zero vs fine-tuned 3D-LLM on ScanQA n-gram metrics | Agent3D-Zero · ScanQA val | BLEU-4 / EM: CIDEr 69.4 -> 71.8; BLEU-4 12.0 -> 4.4 | Mixed |
| use GPT-4V with iteratively chosen views instead of single-image LLaVA | zero-shot VLM · ScanQA val | CIDEr: 5.7 -> 71.8 | Improved |
Later papers that compared with this method
1 later accepted paper reports Agent3D-Zero as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Agentic Collaborative Cognition for Zero-Shot 3D Understanding (ECCV 2026 · better on 2 of 2 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
- IR3D-Bench: Evaluating Vision-Language Model Scene Understanding as Agentic Inverse Rendering (NeurIPS 2025 · similarity 0.91)
- GPT4Scene: Understand 3D Scenes from Videos with Vision-Language Models (ICLR 2026 · similarity 0.90)
- Abstract 3D Perception for Spatial Intelligence in Vision-Language Models (CVPR 2026 · similarity 0.89)
- Pursuing Minimal Sufficiency in Spatial Reasoning (ICLR 2026 · similarity 0.89)
- An Embodied Generalist Agent in 3D World (ICML 2024 · similarity 0.89)
- EmbodiedScan: A Holistic Multi-Modal 3D Perception Suite Towards Embodied AI (CVPR 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.
