Wenqi Marshall Guo

I am an MS student at the University of British Columbia. My work focuses on generative models and alignment. In image generation, I study negative guidance for few-step models and examine how aesthetic alignment can reinforce beauty bias or narrow artistic expression. I also investigate alignment in language-model responses, including the risks of overly cautious answers to health-related questions. Alongside these topics, I have worked on efficient search for graph edit distance, comparing metric search trees with brute-force verification. These projects bring together methods for generative systems and critical questions about the behavior and effects of AI systems. I aim to understand both how model outputs can be guided and how alignment choices affect creative expression and user safety. I share research and software through public repositories.