My research sits at the intersection of machine learning, social psychology, and human-computer interaction.
I focus on three main directions:
- Exploring human-like intelligence in modern AI models (e.g., LLMs and VLMs), such as cognitive biases that mirror human social cognition
- Building practical AI tools to accelerate social science research and human behavior understanding
- Addressing the societal aspects of AI, including privacy protection, security, and sustainable technology innovation
Featured Works
Xuan Liu, Haoyang Shang, Haojian Jin. "CoBRA: Programming Cognitive Bias in Social Agents using Classic Social Science Experiments" CHI 2026 [Code]
Xuan Liu, Jie Zhang, Song Guo, Haoyang Shang, Chengxu Yang, Quanyan Zhu. "Exploring Prosocial Irrationality for LLM Agents: A Social Cognition View" ICLR 2025 [Code]
Xuan Liu, Siqi Cai, Qihua Zhou, Song Guo, Ruibin Li, Kaiwei Lin. "Mjölnir: Breaking the Shield of Perturbation-Protected Gradients via Adaptive Diffusion" AAAI 2025 [Code]
Xuan Liu, Siqi Cai, Lin Li, Rui Zhang, Song Guo. "MGIA: Mutual Gradient Inversion Attack in Multi-Modal Federated Learning" AAAI 2023 [Code]
Chinese Patent CN 109589733 B - An environmentally friendly outdoor dust reduction device using solar thermal energy.
GreenOxy: A Living Air Purifier - Combining advanced technology with natural elements to bring vitality and serenity to daily life.