Research map / Vision-language and multimodal / Multimodal LLMs
Reasoning Path and Latent State Analysis for Multi-view Visual Spatial Reasoning: A Cognitive Science Perspective
Qiyao Xue, Haoming Wang, Weichen Liu, Shiqi Wang, Yuyang Wu, Wei Gao
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 |
|---|---|---|---|
| use object-centric instead of view-centric viewpoint pattern | VLMs · ReMindView-Bench | Avg accuracy: 40.6 -> 37.0 (GPT-4o) | Improved |
| scale Qwen2.5-VL from 3B to 32B | Qwen2.5-VL · ReMindView-Bench | Avg accuracy: 32.2 -> 37.5 | Improved |
| scale InternVL3.5 from 8B to 38B | InternVL3.5 · ReMindView-Bench | Avg accuracy: 25.3 -> 32.6 | Improved |
| scale InternVL3.5 from 2B to 8B | InternVL3.5 · ReMindView-Bench | Avg accuracy: 31.6 -> 25.3 | Hurt |
| use spatial-specialized models vs general open-source | SpatialVLM / SpaceQwen2.5-VL · ReMindView-Bench | Avg accuracy: 31.6 and 28.4 vs 33.8 (LLaVA-OneVision-7B) | No clear effect |
| require cross-frame reasoning vs within-frame | VLMs · ReMindView-Bench | accuracy | Hurt |
| perspective-changing queries vs self-perspective | VLMs · ReMindView-Bench | accuracy: drop >4% | Hurt |
| increase number of objects in scene | VLMs · ReMindView-Bench | accuracy | Hurt |
| progress through reasoning phases (phase 1 to 3) | Qwen2.5-VL-32B · ReMindView-Bench | judge correctness (%): 81.9 -> 38.8 | Hurt |
| scale Qwen2.5-VL from 3B to 32B (self-consistency) | Qwen2.5-VL · ReMindView-Bench | self-consistency overall (%): 48.6 -> 68.4 | Improved |
| correct vs incorrect answers on self-consistency | Qwen2.5-VL-32B · ReMindView-Bench | self-consistency (%): 54.2 -> 83.6 | Improved |
| increase model size on linear probe loss | Qwen2.5-VL family · ReMindView-Bench | probe cross-entropy | Hurt |
Closest papers by meaning
- 3ViewSense: Spatial and Mental Perspective Reasoning from Orthographic Views in Vision-Language Models (ICML 2026 · similarity 0.93)
- Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Models (NeurIPS 2024 · similarity 0.93)
- Spatial-DISE: A Unified Benchmark for Evaluating Spatial Reasoning in Vision-Language Models (ICLR 2026 · similarity 0.92)
- Why Is Spatial Reasoning Hard for VLMs? An Attention Mechanism Perspective on Focus Areas (ICML 2025 · similarity 0.92)
- SpinBench: Perspective and Rotation as a Lens on Spatial Reasoning in VLMs (ICLR 2026 · similarity 0.92)
- Thinking in Structures: Evaluating Spatial Intelligence in Constraint-Governed Spaces (ICML 2026 · similarity 0.92)
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.
