MT-EditFlow: Reinforcement Learning for Multi-Turn Image Editing with Flow Matching
2026-06-01 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors study how current image editing models struggle when users try to make many edits step-by-step, because early mistakes can cause problems later on. They propose MT-EditFlow, a new approach that uses reinforcement learning to better handle multi-turn editing by carefully combining rewards from each step. Their method helps the model plan edits more effectively across the whole sequence, making it better at completing multiple edits without errors. Experiments show their system outperforms existing open-source models and reduces common issues like error buildup.
instruction-based image editingmulti-turn editingreinforcement learningreward signalexposure biasflow matchingadvantage fusionVLM reasoningGRPONFT
Authors
Jiahui Huang, Yasi Zhang, Tianyu Chen, Shu Wang, Jianwen Xie, Oscar Leong, Mingyuan Zhou, Nanzhu Wang, Ying Nian Wu
Abstract
Recent breakthroughs in instruction-based image editing have captured significant attention, as models are now capable of handling real-world editing demands with the practicality required by everyday users. However, editing models trained primarily for single-turn edits often break down in multi-turn editing--the natural interactive setting where a user iteratively refines an image based on the model's own previous outputs. This failure stems from the all-or-nothing requirement, where a single failed turn compromises the entire sequence, and error propagation, where exposure bias leads to compounding editing errors. To address these challenges, we introduce MT-EditFlow, a flow-matching reinforcement learning framework designed to optimize reward signals for sequential image editing. MT-EditFlow integrates a multi-turn perspective with a multi-reward formulation to provide a unified structure applicable to both GRPO and NFT-based reinforcement learning methods. We systematically analyze and optimize the reward signal by investigating effective scoring strategies for turn-level aggregation, VLM reasoning modes to trade off reward bias and variance, and advantage fusion levels to prevent reward hacking. Our findings reveal that broadcasting the aggregated advantage across the entire editing trajectory effectively bridges the gap between local planning and global multi-turn task success. Extensive experiments demonstrate that MT-EditFlow significantly improves performance across diverse base models. Notably, it boosts FLUX.1-Kontext-dev by 6.85 points in turn-3 overall performance, surpassing state-of-the-art open-source models such as Qwen-Image-Edit. By maintaining high marginal success rates and reducing exposure bias, MT-EditFlow provides a foundation for more reliable and natural human-AI collaboration in visual content creation.