Research map / Vision-language and multimodal / Multimodal LLMs
CAPTURE: Evaluating Spatial Reasoning in Vision Language Models via Occluded Object Counting
Atin Pothiraj, Elias Stengel-Eskin, Jaemin Cho, Mohit Bansal
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 |
|---|---|---|---|
| occlude part of the pattern | avg of 6 VLMs · CAPTURe real | sMAPE: 21.95 -> 27.59 | Hurt |
| occlude part of the pattern | avg of 6 VLMs · CAPTURe synthetic | sMAPE: 11.89 -> 15.64 | Hurt |
| occlude part of the pattern | CountGD · CAPTURe real (FSC-147 test subset) | sMAPE: 3.15 -> 10.34 | Hurt |
| feed CountGD visible-object count into VLM prompt | VLMs · CAPTURe real (occluded) | sMAPE | Improved |
| add oracle coordinates of all objects (ablation) | GPT-4o · CAPTURe real (occluded) | sMAPE: 14.75 -> 2.93 | Improved |
| add oracle coordinates of visible objects (ablation) | Qwen2VL · CAPTURe real (occluded) | sMAPE: 29.33 -> 17.70 | Improved |
| add inpainting-predicted occluded information (ablation) | GPT-4o · CAPTURe real (occluded) | sMAPE: 14.75 -> 15.89 | Hurt |
| add inpainting-predicted occluded information (ablation) | Qwen2VL · CAPTURe real (occluded) | sMAPE: 29.33 -> 22.64 | Improved |
| add chain-of-thought prompting (ablation) | Qwen2VL · CAPTURe real (occluded) | sMAPE: 29.33 -> 31.57 | Hurt |
| add chain-of-thought prompting (ablation) | Qwen2VL · CAPTURe synthetic (occluded) | sMAPE: 11.74 -> 37.81 | Hurt |
| add chain-of-thought prompting (ablation) | GPT-4o · CAPTURe synthetic (occluded) | sMAPE: 9.71 -> 7.73 | Improved |
| add chain-of-thought prompting (ablation) | GPT-4o · CAPTURe real (occluded) | sMAPE: 14.75 -> 14.94 | No clear effect |
Closest papers by meaning
- Why Is Spatial Reasoning Hard for VLMs? An Attention Mechanism Perspective on Focus Areas (ICML 2025 · similarity 0.90)
- UniBench: Visual Reasoning Requires Rethinking Vision-Language Beyond Scaling (NeurIPS 2024 · similarity 0.90)
- Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Models (NeurIPS 2024 · similarity 0.90)
- SpatiaLab: Can Vision–Language Models Perform Spatial Reasoning in the Wild? (ICLR 2026 · similarity 0.89)
- BOP-ASK: Object-Interaction Reasoning for Vision-Language Models (CVPR 2026 · similarity 0.89)
- From Flatland to Space: Teaching Vision-Language Models to Perceive and Reason in 3D (NeurIPS 2025 · 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.
