Yizhou Chi

CHI YIZHOU 池一舟

I am a PhD candidate @ University of Cambridge. I am supervised by Andreas Vlachos.

Previously, I obtained my M.S. in EECS and B.S. in Computer Science and Cognitive Science from UC Berkeley, where I was advised by Dan Klein.

My research interest centers around facilitating meaningful and safe communication between humans and computers through language interaction. This entails developing AI systems capable of comprehending both implicit and explicit meanings conveyed by humans, maintaining long-term coherence in conversations, and continuously improving through self-correction and self-examination.

Publications

• SELA: Tree-Search Enhanced LLM Agents for Automated Machine Learning
  Yizhou Chi, Yizhang Lin, Sirui Hong, Duyi Pan, Yaying Fei, Guanghao Mei, Bangbang Liu, Tianqi Pang, Jacky Kwok, Ceyao Zhang, Bang Liu, Chenglin Wu
  We introduce SELA, an innovative agent-based system that leverages Monte Carlo Tree Search (MCTS) to optimize the AutoML process. By representing pipeline configurations as trees, our framework enables agents to conduct experiments intelligently and iteratively refine their strategies, facilitating a more effective exploration of the machine learning solution space.
  [paper]

• THOUGHTSCULPT: Reasoning with Intermediate Revision and Search
  Yizhou Chi, Kevin Yang, Dan Klein
  NAACL 2024 Findings
  We present THOUGHTSCULPT, a general reasoning and search method for tasks with outputs that can be decomposed into components. THOUGHTSCULPT explores a search tree of potential solutions using Monte Carlo Tree Search (MCTS), building solutions one action at a time and evaluating according to any domain-specific heuristic.
  [paper] [code]

• CLARINET: Augmenting Language Models to Ask Clarification Questions for Retrieval
  Yizhou Chi, Jessy Lin, Kevin Lin, Dan Klein
  We present CLARINET, a system that asks informative clarification questions in information retrieval. By augmenting a large language model to condition on the retrieval distribution, CLARINET outperforms traditional heuristics and vanilla-prompted LLMs.
  [paper]

• Feature Selection of High Dimensional Data by Adaptive Potential Particle Swarm Optimization
  Xingyue Huang, Yizhou Chi, Yu Zhou
  2019 IEEE Congress on Evolutionary Computation (CEC)
  We proposed an improved algorithm on Particle Swarm Optimization, capable of classifying high-dimensional data with small training samples; The proposed algorithm makes use of ReliefF to reduce irrelevant features and follows an adapted way to select cut-points based on the feature size of the dataset to achieve higher average accuracy compared to the existing PSO algorithms.
  [paper]

Projects

• Among Agents
  The 4th Wordplay: When Language Meets Games @ ACL 2024
  We introduce AmongAgents, a text-based game environment that mirrors the dynamics of “Among Us” to analyze the behavior of simulated language agents. Our work demonstrates that large language models can effectively grasp game rules and make decisions in socially driven scenarios with incomplete information.
  [paper]

• Interactive Flock Simulator
  [demo site] [github]

• An Empirical Study on Two-child Policy in China Based on Statistical Analysis and Machine Learning
  Yizhou Chi, Xingyue Huang, Yu Zhou
  [paper]