Zhixuan Liu

I am an MSR student working on Generative AI in the Robotics Institute at Carnegie Mellon University, advised by Prof. Jean Oh at roBot Intelligence Group (BIG). I also work closely with Dr. Ji Zhang and Prof. Soonmin Hwang. Before joining CMU, I received my Bachelor's degree in Computer Science and Engineering from The Chinese University of Hong Kong, Shenzhen in 2022.

Email / CV / LinkedIn / Google Scholar / GitHub

Research Interests

My research interests lie in the area of generative models and their application in interactive systems. During my master's thesis at CMU, I focused on addressing cultural biases in text-to-image generative models. Currently, my work involves applying generative models to create simulation environments, particularly in the context of indoor scenes, where simulated robots can navigate and interact in the environment.

I am seeking PhD positions beginning Fall 2024!

Publications

	SCoFT: Self-Contrastive Fine-Tuning for Equitable Image Generation Zhixuan Liu, Peter Schaldenbrand, Beverley-Claire Okogwu, Wenxuan Peng, Youngsik Yun, Andrew Hundt, Jihie Kim, Jean Oh IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2024 Paper / Project Page / Code SCoFT leverages the model's intrinsic biases to refine itself, for the purpose of shifting away from misrepresentations of a culture and achieve equitable image generation.
	Towards Equitable Representation in Text-to-Image Synthesis Models with the Cross-Cultural Understanding Benchmark (CCUB) Dataset Zhixuan Liu, Youeun Shin, Beverley-Claire Okogwu, Youngsik Yun, Lia Coleman, Peter Schaldenbrand, Jihie Kim, Jean Oh AAAI workshop on Creative AI Across Modalities, 2023 Paper / Code Fine-tuning the the text-to-image generative model (Stable Diffusion) and LLM (GPT-3) using our CCUB dataset to achieve culturally-aware text-to-image synthesis.
	StyleCLIPDraw: Coupling Content and Style in Text-to-Drawing Translation Peter Schaldenbrand, Zhixuan Liu, Jean Oh International Joint Conference on Artificial Intelligence (IJCAI), 2022 NeurIPS Workshop on Machine Learning for Creativity and Design, 2021 (Oral) Paper / Oral Presentation / Code / Demo / What's AI on YouTube StyleCLIPDraw is a text-to-drawing synthesis model with artistic control via a given style image and content control via a language description.
	Towards Real-Time Text2Video via CLIP-Guided, Pixel-Level Optimization Peter Schaldenbrand, Zhixuan Liu, Jean Oh NeurIPS Workshop on Machine Learning for Creativity and Design, 2022 Paper / Project Page / Demo / Code An approach to generating videos in real-time based on a series of given language descriptions.
	SongBot: An Interactive Music Generation Robotic System for Non-musicians Learning from a Song Kaiwen Xue, Zhixuan Liu, Jiaying Li, Xiaoqiang Ji, Huihuan Qian IEEE International Conference on Real-time Computing and Robotics (RCAR), 2021 Paper / Video SongBot is an interactive music generation system for the non-musician learners to get inspired from a song.