Rethinking Structure Preservation in Text-Guided Image Editing with Visual Autoregressive Models

2026-03-30

AI Summary

Existing text-guided image editing methods based on visual autoregressive (VAR) models suffer from limitations in spatial localization accuracy and structural consistency. This work proposes a novel framework that analyzes the intermediate feature distributions of VAR models to enable precise editing control. Specifically, it introduces a coarse-to-fine token localization strategy coupled with a structure-aware feature injection mechanism, and further incorporates reinforcement learning to achieve adaptive feature fusion. The proposed approach significantly enhances structural preservation and overall output quality, outperforming state-of-the-art methods on both local and global image editing tasks.

Abstract

Visual autoregressive (VAR) models have recently emerged as a promising family of generative models, enabling a wide range of downstream vision tasks such as text-guided image editing. By shifting the editing paradigm from noise manipulation in diffusion-based methods to token-level operations, VAR-based approaches achieve better background preservation and significantly faster inference. However, existing VAR-based editing methods still face two key challenges: accurately localizing editable tokens and maintaining structural consistency in the edited results. In this work, we propose a novel text-guided image editing framework rooted in an analysis of intermediate feature distributions within VAR models. First, we introduce a coarse-to-fine token localization strategy that can refine editable regions, balancing editing fidelity and background preservation. Second, we analyze the intermediate representations of VAR models and identify structure-related features, by which we design a simple yet effective feature injection mechanism to enhance structural consistency between the edited and source images. Third, we develop a reinforcement learning-based adaptive feature injection scheme that automatically learns scale- and layer-specific injection ratios to jointly optimize editing fidelity and structure preservation. Extensive experiments demonstrate that our method achieves superior structural consistency and editing quality compared with state-of-the-art approaches, across both local and global editing scenarios.

Problem

• text-guided image editing
• visual autoregressive models
• structure preservation
• token localization
• structural consistency

Innovation

• visual autoregressive models
• structure preservation
• token localization
• feature injection
• reinforcement learning