Oriel — AI & technology worth noticing, every day.

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

Paper · https://arxiv.org/abs/2611.06741

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.

First seen by Oriel on .

AI summary

TailSteer is a final-stage training framework that reduces visual drift across 15–20 multi-turn image-editing iterations while preserving foundation models’ existing capabilities.

  • It keeps early denoising dynamics near the base model and applies preservation corrections primarily near the trajectory tail.
  • Corrections are learned from the editor’s own on-policy denoising rollouts using unchanged source content as a preservation reference.
  • TailSteer requires no curated ground-truth editing pairs and avoids retraining on multi-turn datasets or cross-turn rollouts.
  • Experiments on the FLUX, SenseNova, and Qwen families show reduced cumulative drift while largely retaining text-to-image and single-image editing performance.

#image editing #diffusion models #model training #multi-turn interaction

Generated by gpt-5.6-sol. Check the original for details.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained

Points

Score reported by the source.

  • Hugging Face Papers: 43, latest observation ; +8 over observed period since

Comments

Comment count reported by the source.

  • Hugging Face Papers: 4, latest observation ; +3 over observed period since

Aliases

SourceExternal IDURL
Hugging Face Papers2611.06741https://arxiv.org/abs/2611.06741 Read discussion