More about HKUST
Post-Training for Visual Synthesis: Safe and Powerful Generation
PhD Qualifying Examination Title: "Post-Training for Visual Synthesis: Safe and Powerful Generation" by Mr. Runtao LIU Abstract: Recent advancements in visual generation, like text-to-image (T2I) and text-to-video (T2V), have revolutionized creative workflows and content automation, driving unprecedented growth in research and industry applications. However, the rapid development of these technologies has raised critical concerns about their potential misuse in generating harmful content and their limited alignment with user preferences. In this survey we explore existing post-training strategies designed to enhance the safety and quality of visual synthesis while preserving the generative power of pre-trained models. We focus on two core challenges: (1) mitigating risks of unsafe content generation, and (2) optimizing outputs to better align with human preferences in both image and video domains. By summarizing current progress and identifying underexplored research directions, this survey aims to guide the development of robust, user-aligned visual synthesis systems that responsibly harness the potential of generative AI. Date: Friday, 9 May 2025 Time: 2:00pm - 4:00pm Venue: Room 5504 Lifts 25/26 Committee Members: Dr. Qifeng Chen (Supervisor) Dr. Dan Xu (Chairperson) Dr. May Fung