An automated framework to learn reward models from environmental interactions without human annotations, significantly improving large language model agents' performance in complex, multi-step decision-making tasks.
A generative neuro-symbolic visual reasoning model that leverages LLMs to iteratively build, reuse, and refine modular code components, enabling efficient transferability, transparency, and strong performance on visual reasoning tasks.
Educations
University of Massachusetts Amherst
MS, Computer Science, 2024.09 - (now),
Tsinghua University
Undergraduate, Information and Computing Science, 2020.09 - 2024.06