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

DevFeed: [TailSteer: Trajectory-Tail Steering for Drift-Resistant Multi-turn Image Editing](<https://devfeed.tech/articles/tailsteer-trajectory-tail-steering-for-drift-resistant-multi-turn-image-editing-118504.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.06741>)

Author: Ruikang Li, Zixuan Chen, Jiarui Wu, Mingde Yao, Yiyuan Zhang, Yujin Wang, Hanming Deng, Lewei Lu, Tianfan Xue

Published: 2026-11-05T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [image editing](<https://devfeed.tech/topics/image-editing.md>), [Image](<https://devfeed.tech/topics/image.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>)

Tags: [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [computer-science](<https://devfeed.tech/tags/computer-science.md>), [computer-vision](<https://devfeed.tech/tags/computer-vision.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cv](<https://devfeed.tech/tags/cs-cv.md>), [denoising](<https://devfeed.tech/tags/denoising.md>), [flux](<https://devfeed.tech/tags/flux.md>), [foundation-models](<https://devfeed.tech/tags/foundation-models.md>), [framework](<https://devfeed.tech/tags/framework.md>), [generative](<https://devfeed.tech/tags/generative.md>), [image](<https://devfeed.tech/tags/image.md>), [image-editing](<https://devfeed.tech/tags/image-editing.md>), [inference](<https://devfeed.tech/tags/inference.md>), [qwen](<https://devfeed.tech/tags/qwen.md>), [robustness](<https://devfeed.tech/tags/robustness.md>), [steering](<https://devfeed.tech/tags/steering.md>), [text-to-image](<https://devfeed.tech/tags/text-to-image.md>), [training](<https://devfeed.tech/tags/training.md>)

## AI overview

TailSteer aims to reduce visual drift that accumulates when edited images are reused across many turns, while retaining the base model's existing editing and generation capabilities. In this final-stage training framework, early denoising stays close to the base model, and learned preservation corrections are concentrated near the trajectory's end. The corrections use the editor's own on-policy rollouts and unchanged source content as a reference, without curated ground-truth editing pairs. Experiments across FLUX, SenseNova, and Qwen report substantially less cumulative drift over 15-20 editing turns, with base performance largely preserved on text-to-image and single-image editing benchmarks.

## Source excerpt

arXiv:2611.06741v1 Announce Type: cross Abstract: Strong single-turn editing performance does not guarantee reliable multi-turn interaction: as edited outputs are repeatedly reused, small unintended changes can accumulate into severe visual drift. This long-term drift is often mitigated through training on multi-turn data or cross-turn rollouts, potentially altering pretrained priors, or through specialized inference-time mechanisms that are difficult to scale to foundation models. We introduce TailSteer, a final-stage training framework that improves multi-turn robustness while retaining the generative and editing capabilities acquired through earlier training. TailSteer follows a simple principle: keep early denoising dynamics close to those of the base model and concentrate preservation corrections near the trajectory tail, where fine-grained deviations can be suppressed. These corrections are learned from the editor's own on-policy denoising rollouts, using unchanged source content as a preservation reference without requiring curated ground-truth editing pairs. Experiments across the FLUX, SenseNova, and Qwen families show that TailSteer substantially reduces cumulative drift over 15-20 editing turns while largely preserving base-model performance on extensive text-to-image and single-image editing benchmarks. These results demonstrate that long-horizon editing robustness can be unlocked in foundation models after TailSteer, rather than by relearning their existing capabilities.