Selected papers include 'Watch the Weights: Unsupervised monitoring and control of fine-tuned LLMs' and others; selected awards include Gold Medal, Fourth Place at International Olympiad in Informatics 2019, First Place at Meta Hacker Cup 2024, Second Place at 46th ICPC World Final, and more.
Research Experience
Worked as a research scientist at Pika, leading the development of Pika 2.0.
Education
Currently a second-year PhD student at Carnegie Mellon University, advised by Aditi Raghunathan; previously studied computer science and math at MIT.
Background
Research Interest: Understanding, improving, and aligning large language models. Focuses on enhancing their robustness, controllability, and alignment with human values.
Miscellany
Side projects include: Is my problem new? (a semantic search demo for competitive programming problems), CP Ideas (generate competitive programming problems with GPT-3), Mosaic Detective (guess the word from an image downscaled), Light Synth (synthesizer, but with balls and whistles), Pitch Trainer (microtonal ear training, in Chinese).