Research map / Vision-language and multimodal / Multimodal LLMs
Visual Spatial Tuning
Rui Yang, ziyu zhu, Yanwei Li, Jingjia Huang, Shen Yan, Siyuan Zhou, Zhe Liu, Xiangtai Li, Shuangye Li, Wenqian Wang, Yi Lin, Hengshuang 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 single-image scene caption data (ablation) | VST-3B SFT · spatial benchmarks | overall improvement: 2.3% | Improved |
| add depth and distance data (ablation) | VST-3B SFT · CVBench-3D | accuracy: 10.4% | Improved |
| add multi-view correspondence data (ablation) | VST-3B SFT · BLINK | accuracy: 1.8% | Improved |
| add multi-image 3D object detection data (ablation) | VST-3B SFT · MMSI-Bench | accuracy: 2.1% | Improved |
| add general video data (ablation) | VST-3B SFT · VSI-Bench | accuracy: 38.4 -> 38.1 | Hurt |
| add VST-video data (ablation) | VST-3B SFT · VSI-Bench | accuracy: 38.4 -> 54.7 | Improved |
| generate CoT conditioned on BEV image (ablation) | Qwen2.5-VL-7B cold start · MMSI-Bench | overall accuracy: 30.0 -> 31.1 | Improved |
| add CoT inference after cold start (ablation) | Qwen2.5-VL-7B cold start · MMSI-Bench | overall accuracy: 33.6 -> 31.7 | Hurt |
| add RL stage after cold start (ablation) | VST cold-start model · MMSI-Bench | overall accuracy: 31.7 -> 35.3 | Improved |
| add FoV unification (ablation) | VST 3D detection · ARKitScenes | AP@15: 40.9 -> 42.7 | Improved |
| use multi-turn instead of single-turn detection data (ablation) | VST 3D detection · SUN RGB-D | AP@15: 26.0 -> 28.5 | Improved |
| use IoU plus F1 reward instead of IoU plus recall (ablation) | VST RL 3D detection · SUN RGB-D | AP: 13.8 -> 24.4 | Improved |
Closest papers by meaning
- SpatialRGPT: Grounded Spatial Reasoning in Vision-Language Models (NeurIPS 2024 · similarity 0.92)
- Expand VSR Benchmark for VLLM to Expertize in Spatial Rules (AAAI 2025 · similarity 0.92)
- Spatial Understanding from Videos: Structured Prompts Meet Simulation Data (NeurIPS 2025 · similarity 0.92)
- InternSpatial: A Comprehensive Dataset for Spatial Reasoning in Vision-Language Models (ICLR 2026 · similarity 0.91)
- Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Models (NeurIPS 2024 · similarity 0.91)
- SpatialLadder: Progressive Training for Spatial Reasoning in Vision-Language Models (ICLR 2026 · similarity 0.91)
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.
