My research focuses on neural-symbolic reasoning, which entails synergistically leveraging the strengths of machine learning systems, specifically large language models (LLMs), with symbolic systems such as knowledge graphs and combinatorial optimization solvers. The goal of my research is to develop a versatile intelligent system that can truly comprehend the semantics of natural languages and possess compositional reasoning capabilities for accurately solving complex tasks.

I have also conducted research in related fields, including efficient sequence generation (non-autoregressive generation and sparse attention), vision-grounded reasoning (object detection and VQA), model compression, and neural PDE solving (AI4Science).