Research map / Vision-language and multimodal / Multimodal LLMs
MM-Spatial: Exploring 3D Spatial Understanding in Multimodal LLMs
Erik Daxberger, Nina Wenzel, David Griffiths, Haiming Gang, Justin Lazarow, Gefen Kohavi, Kai Kang, Marcin Eichner, Yinfei Yang, Afshin Dehghan, Peter Grasch
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 multi-view input (ablation) | MM-Spatial-3B specialist · CA-VQA benchmark (specialist) | accuracy (average): 49.4 -> 51.4 | Improved |
| add chain-of-thought depth reasoning (ablation) | MM-Spatial-3B generalist · CA-VQA benchmark (generalist) | accuracy (average): 47.0 -> 49.1 | Improved |
| add GT depth via tool-use (ablation) | MM-Spatial-3B specialist · CA-VQA benchmark (specialist) | accuracy (average): 49.4 -> 54.5 | Improved |
| add fully encoded GT depth maps (ablation) | MM-Spatial-3B specialist · CA-VQA benchmark (specialist) | accuracy (average): 49.4 -> 49.9 | No clear effect |
| add monocular depth via tool-use (ablation) | MM-Spatial-3B specialist · CA-VQA benchmark (specialist) | accuracy (average): 49.4 -> 49.5 | No clear effect |
| use GT depth tool instead of fully encoded depth (ablation) | MM-Spatial-3B specialist · CA-VQA benchmark (specialist) | accuracy (average): 49.9 -> 54.5 | Improved |
| use CoT instead of monocular depth tool (ablation) | MM-Spatial-3B specialist · CA-VQA benchmark (specialist) | accuracy (average): 49.5 -> 50.8 | Improved |
| add multi-view to GT depth tool (ablation) | MM-Spatial-3B specialist · CA-VQA benchmark (specialist) | accuracy (average): 54.5 -> 55.3 | Improved |
| add image input (vs blind eval) (ablation) | MM-Spatial-3B specialist · CA-VQA benchmark (specialist) | accuracy (average): 24.0 -> 49.4 | Improved |
| add scale augmentation to CA-VQA* (ablation) | MM-Spatial-3B + Depth (Tool; Mon.) · SpatialRGPT-Bench (average) | accuracy: 74.9 -> 77.7 | Improved |
| add OSD to CA-VQA* training data (ablation) | MM-Spatial-3B + Depth (Tool; Mon.) · SpatialRGPT-Bench (average) | accuracy: 74.9 -> 81.0 | Improved |
| add monocular metric depth via tool-use (ablation) | MM-Spatial-3B (CA-VQA* + OSD) · SpatialRGPT-Bench (average) | accuracy: 77.8 -> 81.0 | Improved |
Later papers that compared with this method
1 later accepted paper reports MM-Spatial-3B as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- HiSpatial: Taming Hierarchical 3D Spatial Understanding in Vision-Language Models (CVPR 2026 · better on 1 of 1 shared result)
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
- SD-VLM: Spatial Measuring and Understanding with Depth-Encoded Vision-Language Models (NeurIPS 2025 · similarity 0.92)
- Video-3D LLM: Learning Position-Aware Video Representation for 3D Scene Understanding (CVPR 2025 · similarity 0.91)
- MMScan: A Multi-Modal 3D Scene Dataset with Hierarchical Grounded Language Annotations (NeurIPS 2024 · similarity 0.91)
- Multi-SpatialMLLM: Multi-Frame Spatial Understanding with Multi-Modal Large Language Models (CVPR 2026 · similarity 0.91)
- MMSI-Bench: A Benchmark for Multi-Image Spatial Intelligence (ICLR 2026 · similarity 0.91)
- Spatial-MLLM: Boosting MLLM Capabilities in Visual-based Spatial Intelligence (NeurIPS 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.
