Yizhou Chi

I am currently a 5th year EECS master's student @ UC Berkeley, advised by Dan Klein, and I just graduated with a double major in Computer Science and Cognitive Science in Spring 2023.

My research interest centers around facilitating meaningful 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.

Student Researcher, Berkeley NLP Group, Jun 2022 – present (Advisor: Dan Klein)

• Research on a comprehensive dialogue-based clarification system. The project aims to more efficiently help humans identify and retrieve information and documents by prompting users with generated open-ended clarification questions. By handling the users’ responses using the language model T5, the system updates its belief and narrows down choices from the DPR retriever.
• Research on Large Language Models’ self-revision capabilities. The research seeks to explore the potential of LLMs (GPT3.5 and GPT4) in revising decisions and making improvements. We incorporate a range of tasks including story outline revision, crosswords, and constrained generation to learn how LLMs, by leveraging Monte Carlo Tree Search, can evaluate their previous outputs and produce revised and improved outputs.

Student Researcher, Abbasi Lab, Sept 2021 – Jun 2023 (Advisor: Reza Abbasi)

• Collaborate with the Allen Institute for Brain Science to develop models that make classifications using a dataset of simulated neuron spikes in a mouse’s visual cortex; build a pipeline that seeks to maximize the classification performance with neural networks.
• Seek to reconstruct visual information from neuron spikes using state-of-the-art image reconstruction methods including GANs and diffusion models.

UC Berkeley - EECS Master of Science