| title | MTEdit: Packed Trajectories Beat Rollouts for Multi-Turn Image Editing |
|---|---|
| authors | Houyuan Chen, Fanzhou Wang, Xiangyu Fan, Songlin Yang, Zhitao Yang, Lvmin Zhang, Pei Yang, Hao Zhao, Lei Yang, Anyi Rao |
| date | 2026-10-16 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2610.23751 |
| affiliation | SenseNova |
| upvotes | 29 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| abstract | Multi-turn image editing aims to modify an image according to a sequence of instructions, making it better aligned with human-AI collaborative creation than single-turn editing. However, constructing high-quality multi-turn training data remains challenging, as conventional sequential rollout accumulates errors and introduces visual-quality drift across editing turns. To address this challenge, we introduce MTEdit, a multi-turn image editing dataset built through a pipeline centered on packed trajectory canvas (PTC) generation. By jointly generating multiple editing states within a unified visual context, this pipeline reduces visual-quality drift across turns. Covering both step-wise and context-aware editing scenarios, MTEdit comprises 500K multi-turn image editing trajectories, with up to 8 turns and an average of 5.38 turns per trajectory, making it one of the most extensive datasets for multi-turn image editing. We further introduce MTEdit-Bench, a benchmark designed to evaluate both scenarios by assessing models' ability to follow editing instructions, maintain content consistency, resist visual-quality drift, and effectively use historical context across turns. Using MTEdit, we fine-tune Qwen-Image-Edit-Causal to obtain Qwen-Image-MTEdit. On MTEdit-Bench, we evaluate a broad range of closed-source and open-source image editing models, including Qwen-Image-MTEdit. Experimental results show that Qwen-Image-MTEdit achieves the best performance among the evaluated open-source models, demonstrating the effectiveness of MTEdit in improving models' multi-turn image editing capabilities. |
cite
@misc{chen2026mtedit,
title={MTEdit: Packed Trajectories Beat Rollouts for Multi-Turn Image Editing},
author={Houyuan Chen and Fanzhou Wang and Xiangyu Fan and Songlin Yang and Zhitao Yang and Lvmin Zhang and Pei Yang and Hao Zhao and Lei Yang and Anyi Rao},
year={2026},
eprint={2610.23751},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2610.23751}
}