Portrait of Xujie Si

Xujie Si

Affiliate Member
Canada CIFAR AI Chair
Assistant Professor, University of Toronto, Department of Computer Science
Research Topics
Learning to Program
Reasoning
Representation Learning

Biography

Xujie Si is an assistant professor in the Department of Computer Science, University of Toronto. He is also an affiliate member at Vector Institute and an affiliate member at Mila – Quebec Artificial Intelligence Institute, where he holds a Canada CIFAR AI Chair.

Si received his PhD from the University of Pennsylvania in 2020, his master's degree from Vanderbilt University, and his bachelor's degree (with Honors) from Nankai University.

Si’s research lies at the intersection of programming languages and AI. He is generally interested in developing learning-based techniques to help programmers build better software with less effort, integrating logic programming with differentiable learning systems for interpretable and scalable reasoning, and leveraging programming abstractions for reliable and data-efficient learning.

His work has been honoured with an ACM Special Interest Group on Programming Languages (SIGPLAN) Distinguished Paper Award, as well as highlighted at top conferences on programming languages and machine learning.

Current Students

Postdoctorate - McGill University
Principal supervisor :
PhD - McGill University
PhD - McGill University
Co-supervisor :
Master's Research - McGill University
Master's Research - McGill University

Publications

Novice Type Error Diagnosis with Natural Language Models
Chuqin Geng
Haolin Ye
Yixuan Li
Tianyu Han
Brigitte Pientka
Strong static type systems help programmers eliminate many errors without much burden of supplying type annotations. However, this flexibili… (see more)ty makes it highly non-trivial to diagnose ill-typed programs, especially for novice programmers. Compared to classic constraint solving and optimization-based approaches, the data-driven approach has shown great promise in identifying the root causes of type errors with higher accuracy. Instead of relying on hand-engineered features, this work explores natural language models for type error localization, which can be trained in an end-to-end fashion without requiring any features. We demonstrate that, for novice type error diagnosis, the language model-based approach significantly outperforms the previous state-of-the-art data-driven approach. Specifically, our model could predict type errors correctly 62% of the time, outperforming the state-of-the-art Nate's data-driven model by 11%, in a more rigorous accuracy metric. Furthermore, we also apply structural probes to explain the performance difference between different language models.