DanceOPD: On-Policy Generative Field Distillation
2026-06-25 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionComputation and LanguageMachine Learning
AI summaryⓘ
The authors address the challenge of building a single image generation model that can handle different tasks like creating images from text and editing images either locally or globally. They introduce DanceOPD, a new training method that helps the model learn these tasks separately but within a shared framework, by routing data to specialized parts and training with a simple loss. This approach lets the model combine expert abilities effectively without one task hurting another. Their tests show improved ability to do multiple tasks while keeping overall image quality stable.
image generationtext-to-image (T2I)local editingglobal editingflow-matching modelsgenerative field distillationvelocity fieldclassifier-free guidancemean squared error (MSE)model composition
Authors
Wei Zhou, Xiongwei Zhu, Zelin Xu, Bo Dong, Lixue Gong, Yongyuan Liang, Meng Chu, Leigang Qu, Lingdong Kong, Wei Liu, Tat-Seng Chua
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, we 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. With each capability source defined as a velocity field over the shared flow state space, the student learns from fields queried on its own rollout states to compose expert capabilities. This formulation also absorbs operator-defined fields such as classifier-free guidance. Comprehensive experiments on T2I, editing, realism-field absorption, and CFG absorption show that our approach improves multi-capability composition, strengthening target capabilities while preserving anchor generation quality. We believe this work establishes a practical route for generative field distillation in flow-matching models.