Research map / Vision-language and multimodal / Human and embodied modeling
Human-like Abstract Visual Reasoning via Understanding and Solving Reasoning Loop
Xinwang Chen, Xiuxing Li, Qing Li, Ziyue Zhuang, Yutong Wu, Ziyu Li, Zhuo Wang, Kai Li, Jianye Hao, Xia Wu
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 Understanding Module without gate to the Solving Module baseline (ablation) | USRL · ARC-AGI-1 eval | accuracy (%): 36.5 -> 40.8 | Improved |
| add gating that modulates rule representation by solution state (ablation) | USRL · ARC-AGI-1 eval | accuracy (%): 40.8 -> 43.2 | Improved |
| add auxiliary contrastive loss for rule-representation isomorphism (ablation) | USRL · ARC-AGI-1 eval | accuracy (%): 43.2 -> 46.4 | Improved |
| add stochastic halting during training (ablation) | USRL · ARC-AGI-1 eval | accuracy (%): 46.4 -> 47.2 | Improved |
| use adaptive reasoning halting at inference instead of fixed steps (ablation) | USRL · ARC-AGI-1 eval | accuracy (%): 47.2 -> 47.0 | No clear effect |
| reduce maximum inference reasoning steps below training depth of 16 (ablation) | USRL · ARC-AGI-1 eval | accuracy (%): 47.0 -> 11.4 (Tmax=8) | Hurt |
| increase maximum inference steps from 16 to 20 (ablation) | USRL · ARC-AGI-1 eval | accuracy (%): 47.0 -> 47.1 | No clear effect |
| replace fixed puzzle-embedding tables with example-derived recurrent understanding module | USRL vs TRM · ARC-AGI-1 eval | accuracy (%): 44.6 -> 47.2 | Improved |
Closest papers by meaning
- Think Visually, Reason Textually: Vision-Language Synergy in Abstract Reasoning (CVPR 2026 · similarity 0.90)
- ARC Is a Vision Problem! (CVPR 2026 · similarity 0.89)
- CogFlow: Bridging Perception and Reasoning through Knowledge Internalization for Visual Mathematical Problem Solving (ICLR 2026 · similarity 0.88)
- Beyond Task-Specific Reasoning: A Unified Conditional Generative Framework for Abstract Visual Reasoning (ICML 2025 · similarity 0.87)
- VGR: Visual Grounded Reasoning (ICLR 2026 · similarity 0.87)
- CodeDance: A Dynamic Tool-integrated MLLM for Executable Visual Reasoning (CVPR 2026 · similarity 0.87)
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.
