About me

I am a Ph.D. student in AI at HKUST (Guangzhou), advised by Prof. Jun Wang. Before this I trained in statistics and machine learning at Warwick and UCL.

Education

• Ph.D. in Artificial Intelligence, HKUST (Guangzhou), 2024 - Present
• M.Sc. in Machine Learning, University College London (UCL), 2022 - 2023
• B.Sc. in Mathematics and Statistics, University of Warwick, 2019 - 2022

Research Interests

I work on active decision-making in modern AI systems: active information acquisition (what to observe) and active intervention (what to do). The shared challenge—information is incomplete, observations are costly, and not every action is worth taking—shapes most of the problems I find interesting. My recent work approaches these questions through Bayesian experimental design and causal intervention, applied to vision-language models, multi-agent reinforcement learning, and LLM agent reasoning.

I care equally about whether the formalizations translate into systems that work outside the benchmark—where information has real cost, decisions have real consequences, and efficiency is the difference between a method that ships and one that doesn’t.

I’m particularly open to collaboration on real-world applications where information has cost and decisions have consequence—such as finance and clinical decision support—and on technical directions including belief representation, uncertainty handling, and information routing in LLM-based agents.

Keywords: LLM agents · decision theory · reinforcement learning · causal inference · Bayesian methods

Publications

• The Perceptual Bandwidth Bottleneck in Vision-Language Models: Active Visual Reasoning via Sequential Experimental Design [Code] [Slides]
  A. Liu*, Z. Gong*, Y. Song, Y. Chen, X. Liu, H. Lu, K. Zhang, C. Wei, J. Wang
  ICML 2026

• A Principle of Targeted Intervention for Multi-Agent Reinforcement Learning [Code]
  A. Liu*, J. Wang*, S. Kaski, J. Wang, M. Yang
  NeurIPS 2025

• Causal Sufficiency and Necessity Improves Chain-of-Thought Reasoning
  X. Yu, Z. Wang, L. Yang, H. Li, A. Liu, X. Xue, J. Wang, M. Yang
  NeurIPS 2025

• OpenR: An open source framework for advanced reasoning with large language models [Code]
  J. Wang, M. Fang, Z. Wan, M. Wen, J. Zhu, A. Liu, Z. Gong, Y. Song, L. Chen, et al.
  arXiv preprint

• Attaining Human’s Desirable Outcomes in Human-AI Interaction via Structural Causal Games
  A. Liu, J. Wang, H. Li, X. Chen, J. Wang, S. Kaski, M. Yang
  ICML 2024 Workshop on Humans-Algs-Society

* Equal contribution