Summary
DanceOPD (arXiv:2606.27377) is a paper by Wei Zhou, Xiongwei Zhu, Tat-Seng Chua and colleagues proposing an on-policy generative field distillation framework for flow-matching image generation models. Modern image generators must unify diverse capabilities—text-to-image (T2I) synthesis, local editing, and global editing—but these abilities are rarely aligned and often conflict: editing degrades T2I quality, and global and local editing interfere with each other. DanceOPD addresses this by defining each capability source as a velocity field over a shared flow state space. At training time, each sample is routed to a single capability field, a low-noise student-induced state is queried, and the model is trained with a simple velocity MSE objective. Because students learn from fields queried on their own rollout states, expert capabilities compose naturally, and the formulation can also absorb operator-defined fields such as classifier-free guidance (CFG). Experiments across T2I, editing, realism field absorption, and CFG absorption show improved multi-capability composition while preserving anchor generation quality.
Paper Overview
Research areas: Computer Vision, Computation and Language, Machine Learning
Authors: Wei Zhou, Xiongwei Zhu, Zelin Xu, Bo Dong, Lixue Gong, Yongyuan Liang, Meng Chu, Leigang Qu, Lingdong Kong, Wei Liu, Tat-Seng Chua
Published: 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 for flow-matching models that:
- routes each sample to one capability field,
- queries one low-noise student-induced state, and
- trains with a simple velocity MSE objective.
Each capability source is defined as a velocity field over the shared flow state space, and the student learns from fields queried on its own rollout states in order to compose expert capabilities. This formulation also absorbs operator-defined fields, such as classifier-free guidance (CFG).
Comprehensive experiments on T2I, editing, realism field absorption, and CFG absorption show that the method improves multi-capability composition, enhancing target capabilities while maintaining anchor generation quality.
---
*Auto-collected on 2026-06-27*
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/178208197