ACEdit: Stable Multi-Turn Image Editing via Asymmetric Conditioning

One-line takeaway

ACEdit achieves stable multi-turn image editing via asymmetric conditioning and distillation, reducing quality degradation over long sequences.

ACEdit通过自生成条件与教师模型蒸馏，实现了多轮图像编辑中的稳定性，显著减少编辑过程中的质量退化。

Key points

ACEdit addresses the degradation of image editing quality over multiple turns by introducing asymmetric conditioning and a teacher-student distillation framework. The student model edits its own outputs sequentially, while a teacher model, anchored to the original image, provides reliable supervision via compound instructions. This setup stabilizes the distribution shift during multi-turn editing, preserving quality across long sequences.

• ACEdit introduces asymmetric conditioning to stabilize multi-turn image editing.
• Self-generated editing trajectories are supervised via a teacher model to avoid direct supervision bias.
• The teacher remains anchored to the original image for consistent and reliable guidance.
• Experiments show significant reduction in quality degradation over long editing sequences.

Written only from the official arXiv abstract by an open-source model, then checked automatically (every number and name must appear in the abstract). The paper is authoritative.