Probing Visual Planning in Image Editing Models

2026-04-23

AI Summary

Current image editing models rely on inefficient step-by-step generation in visual planning, struggling to emulate human-like spatial reasoning. This work proposes the “Editing as Reasoning” (EAR) paradigm, reframing visual planning as a single-step image transformation. To support systematic evaluation of both pixel fidelity and logical correctness, the authors introduce the AMAZE dataset, which effectively disentangles visual perception from reasoning capabilities. Leveraging both autoregressive and diffusion architectures, models fine-tuned on abstract mazes and the N-Queens problem demonstrate notable zero-shot generalization to larger instances and out-of-distribution geometric structures. Although still far less efficient than human reasoning, these results validate the feasibility and potential of the EAR paradigm for integrating visual editing with structured reasoning.

Technology Category

• Computer Vision: Visual Reasoning & Symbolic Representations
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Planning, Routing, and Scheduling: Model-Based Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Visual planning represents a crucial facet of human intelligence, especially in tasks that require complex spatial reasoning and navigation. Yet, in machine learning, this inherently visual problem is often tackled through a verbal-centric lens. While recent research demonstrates the promise of fully visual approaches, they suffer from significant computational inefficiency due to the step-by-step planning-by-generation paradigm. In this work, we present EAR, an editing-as-reasoning paradigm that reformulates visual planning as a single-step image transformation. To isolate intrinsic reasoning from visual recognition, we employ abstract puzzles as probing tasks and introduce AMAZE, a procedurally generated dataset that features the classical Maze and Queen problems, covering distinct, complementary forms of visual planning. The abstract nature of AMAZE also facilitates automatic evaluation of autoregressive and diffusion-based models in terms of both pixel-wise fidelity and logical validity. We assess leading proprietary and open-source editing models. The results show that they all struggle in the zero-shot setting, finetuning on basic scales enables remarkable generalization to larger in-domain scales and out-of-domain scales and geometries. However, our best model that runs on high-end hardware fails to match the zero-shot efficiency of human solvers, highlighting a persistent gap in neural visual reasoning.

Problem

• visual planning
• image editing models
• computational inefficiency
• zero-shot reasoning
• neural visual reasoning

Innovation

• visual planning
• editing-as-reasoning
• AMAZE dataset
• abstract reasoning
• image transformation