
WL
Wenjin Liu, Fayuan Ke, Yue Lu, Zhe Cui, Anh Tuan Luu, Haoran Luo
· 1 min read
ResearcharXiv cs.CV
AdaPilot: Towards Scene-Adaptive Policy Learning for Cross-Generator Text-to-Image Quality Optimization
arXiv:2609.29517v1 Announce Type: new
Abstract: Existing methods for improving text-to-image generation quality have progressed from generator fine-tuning and prompt optimization to reinforcement learning with multi-turn visual feedback. However, existing strategies are deeply coupled with specific generators and tasks, and the learned capabilities are difficult to generalize into a universal quality optimization policy. Therefore, we propose AdaPilot, which learns a scene-adaptive, cross-generator transferable quality optimization policy by formulating multi-turn image generation as a Markov Decision Process (MDP) and optimizing it via end-to-end reinforcement learning. Specifically, AdaPilot decouples the policy from generator internals to enable cross-generator transfer, introduces scene-aware rewards that adaptively align quality assessment dimensions with task semantics, and employs process-level rewards to model the evolution trajectory of image quality. Experimental results show AdaPilot outperforms baselines in generation quality and generalization. Separate cross-generator evaluations further show that a single policy transfers zero-shot to unseen generators while maintaining positive average gains across all evaluated generators. Our project is available at https://github.com/QwenQing/Ada_pilot.
Original source
This story was published by arXiv cs.CV and written by Wenjin Liu, Fayuan Ke, Yue Lu, Zhe Cui, Anh Tuan Luu, Haoran Luo. SyncAI.news shows a preview; the complete article is on the publisher's site.
Read the full story on arxiv.org


