Ruikang Li, Zixuan Chen, Jiarui Wu, Mingde Yao, Yiyuan Zhang, Yujin Wang, Hanming Deng, Lewei Lu (MMLab-CUHK) · Drift-Resistant Editing

TailSteer: Trajectory-Tail Steering for Drift-Resistant Multi-turn Image Editing

TailSteer is a final-stage training framework that reduces visual drift when an image editor's outputs are repeatedly edited over multiple turns. It keeps early denoising dynamics close to the base model and applies preservation corrections near the trajectory tail, learned from the editor's own on-policy rollouts without curated editing pairs. Experiments on FLUX, SenseNova and Qwen families report substantially reduced cumulative drift over 15-20 editing turns while largely preserving base-model performance on text-to-image and single-image editing benchmarks.