I’m a Member of Technical Staff at Cohere, working on foundation models for embeddings and search. I obtained my Ph.D. from the University of Waterloo, supervised by Professor Jimmy Lin. Before that, I received my M.Sc. in Computer Science from University of Toronto under the supervision of Danijar Hafner and Professor Jimmy Ba. I completed my undergraduate study at Sun Yat-Sen University advised by Professor Weishi Zheng.

News

• 2024-09-26: New paper alert! Our NEST paper has been accepted to NeurIPS 2024! The OSS version will be released soon.
• 2024-09-21: New paper alert! Our Document Screenshot Embedding paper has been accepted to EMNLP 2024. We directly encode document screenshots into vectors using visual-LLM for semantic search.
• 2024-08-20: I join the search and embedding team at Cohere as a Member of Technical Staff!
• 2024-08-07: I pass my Ph.D. defence from the University of Waterloo!
• 2024-05-30: New paper alert! Check out our latest work NEST on enchancing factuality and attribution of LLMs as well as discussions on X.
• 2023-10-09: New paper alert! Our paper How to Train Your DRAGON: Diverse Augmentation Towards Generalizable Dense Retrieval has been accepted to EMNLP 2023! Check out this super robust and easy-to-use dense retriever here and try it out yourself!
• 2023-09-18: I start my internship at Meta FAIR Lab as a Research Scientist intern hosted by Victoria Lin.
• 2023-08-22: I’m awarded with the Waterloo Apple PhD Fellowship in Data Science and Machine Learning.
• 2023-06-22: I start my internship at Google Research as a Student Researcher under the supervision of Honglei Zhuang
• 2023-05-02: Our paper CITADEL is accepted in ACL 2023! Check out this efficient multi-vector retriever which i about 40x faster than ColBERT-v2 on GPUs.
• 2022-02-13: New paper is out! SLIM manages to reduce the latency and storage of ColBERT while being fully compatible with Pyserini (Lucene-based). Codes will be released soon!
• 2022-02-13: New paper is out! We find that adding contextualized late interaction could be helpful for cross-encoders on out-of-domain generalization. Check out the paper and some discussion on twitter.