English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

DanceOPD: On-Policy Generative Field Distillation for Unified Image Generation

Forum topic · 小凯 · 2026-06-27

Summary

DanceOPD is a paper by Wei Zhou, Xiongwei Zhu, and Zelin Xu (arXiv:2606.27377, posted June 27, 2026) that proposes an on-policy generative field distillation framework for modern image generation. The work addresses a key challenge in training unified image generation models: combining text-to-image (T2I) synthesis, local editing, and global editing in a single model. These capabilities are rarely naturally aligned and often conflict with one another—editing tends to degrade T2I performance, while global and local editing interfere with each other. DanceOPD tackles this by ensuring each capability learns from the model's own generated (on-policy) distribution, which avoids misalignment and capability conflicts. The Chinese forum post on zhichai.net summarizes the paper's motivation and approach, highlighting how on-policy distillation enables effective composition of multiple image generation capabilities in one model.

Paper Overview

Field: NLP Authors: Wei Zhou, Xiongwei Zhu, Zelin Xu Posted: 2026-06-27 arXiv: 2606.27377

Abstract

Modern image generation demands a single model that unifies diverse capabilities, including text-to-image (T2I), local editing, and global editing. However, these capabilities are rarely naturally aligned and often conflict. For instance, editing tends to degrade T2I performance, while global and local editing interfere with each other. Consequently, effectively composing these capabilities has become a central challenge for image generation model training.

To tackle this, the authors introduce DanceOPD, an on-policy generative field distillation framework. The core idea is to ensure that each capability learns from the model's own generated distribution, rather than from an off-policy or external source. This on-policy design avoids misalignment and prevents the different capabilities from conflicting with each other during training.

Key Points

  • Unifies three capabilities in one model: text-to-image generation, local editing, and global editing.
  • Identifies mutual interference: editing degrades T2I quality; global and local editing conflict with each other.
  • Proposes on-policy generative field distillation so each capability is learned from the model's own output distribution.
> Full details are available in the paper: https://arxiv.org/abs/2606.27377

*Auto-collected on 2026-06-27*

Tags

#image-generation#diffusion-models#knowledge-distillation#text-to-image#on-policy-learning#paper#arxiv

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178208163