REASONEDIT: Towards Reasoning-Enhanced Image Editing Models

2025-11-27

AI Summary

Existing image editing models typically freeze parameters of multimodal large language models (MLLMs), severely limiting their reasoning capabilities and hindering accurate interpretation of abstract editing instructions. To address this, we propose a reasoning-enhanced image editing framework that introduces, for the first time in this domain, a closed-loop “reason–edit–reflect” mechanism. Our approach unfreezes MLLM parameters to enable world-knowledge-guided reasoning, integrates end-to-end optimization with a diffusion-based decoder, and equips the model with capabilities for error detection, autonomous correction, and termination judgment. Evaluated on ImgEdit, GEdit, and Kris benchmarks, our method achieves improvements of 4.3%, 4.7%, and 8.2%, respectively—substantially outperforming leading open-source alternatives. This work establishes a new paradigm for controllable, reasoning-driven image editing.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Diffusion Models for Vision
• Knowledge Representation and Reasoning: Reasoning with Beliefs

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Recent advances in image editing models have shown remarkable progress. A common architectural design couples a multimodal large language model (MLLM) encoder with a diffusion decoder, as seen in systems such as Step1X-Edit and Qwen-Image-Edit, where the MLLM encodes both the reference image and the instruction but remains frozen during training. In this work, we demonstrate that unlocking the reasoning capabilities of MLLM can further push the boundaries of editing models. Specifically, we explore two reasoning mechanisms, thinking and reflection, which enhance instruction understanding and editing accuracy. Based on that, our proposed framework enables image editing in a thinking-editing-reflection loop: the thinking mechanism leverages the world knowledge of MLLM to interpret abstract instructions, while the reflection reviews editing results, automatically corrects unintended manipulations, and identifies the stopping round. Extensive experiments demonstrate that our reasoning approach achieves significant performance gains, with improvements of ImgEdit (+4.3%), GEdit (+4.7%), and Kris (+8.2%) when initializing our DiT from the Step1X-Edit (ReasonEdit-S), and also outperforms previous open-source methods on both GEdit and Kris when integrated with Qwen-Image-Edit (ReasonEdit-Q).

Problem

• Enhances image editing with reasoning mechanisms
• Improves instruction understanding and editing accuracy
• Enables iterative thinking-editing-reflection loop

Innovation

• Unlocking MLLM reasoning for image editing enhancement
• Implementing thinking-reflection loop for instruction interpretation
• Automatically correcting edits and determining optimal stopping point