terryyz.github.io

I am getting my PhD at CSIRO's Data61 and Monash University. I am currently leading the next generation of Code LLMs at BigCode while collaborating with various entities.

My research focuses on code intelligence, particularly function calling, program reasoning, and code generation. The long-term goal is to build AGI with code/executable languages.

I am open to collaboration. Feel free to contact me if you are interested in working together.

My Working Principles

I typically work at least 10 hours a day, unless I decide to take a break.
Choose wisely and commit wholeheartedly.
Paper is cheap; they will be published eventually.
I prioritize long-term research impact over quick wins in publications.
Any upcoming research I lead will require significant effort and can often be split into multiple papers (though I choose not to).
I am a generalist, always open to new opportunities, depending on my bandwidth.

Why Code Intelligence?

Software is the backbone of all advanced technology, making code the bedrock of innovation. [1, 2, 3]
Code's structured and executable nature allows for precise control and automation. [1, 4]
Code is the most potent tool for AI to interact with and manipulate the world. [1, ???]
Code intelligence can dramatically enhance problem-solving capabilities across various domains. [1, ???]
Advancements in code intelligence lead to more efficient, secure, and scalable solutions. [5, 6, 7]

Fun Facts

I started my first research project in Quantum Computing, not Computer Science.
During my undergraduate years, I explored a wide range of NLP topics, including but not limited to: Commonsense Reasoning, Multimodal Learning, Machine Translation, and Semantic Parsing.
In a previous life, I was a semiprofessional swimmer, training for 10 years.
I'm an amateur pianist, playing for 8 years.
Before high school, I had traveled to most European countries.

Recent Publications

BigCode Team, “BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructions ”, Manuscript, 2024. [Github]

BigCode Team, “OctoPack: Instruction Tuning Code Large Language Models”, International Conference on Learning Representations (ICLR), 2024. [Github]

Terry Yue Zhuo, “ICE-Score: Instructing Large Language Models to Evaluate Code”, Findings of European Chapter of the Association for Computational Linguistics (EACL), 2024. [Github]

BigCode Team, “Astraios: Parameter-Efficient Instruction Tuning Code Large Language Models”, Manuscript, 2024.

BigCode Team, “StarCoder: May The Source Be With You!”, Transactions on Machine Learning Research (TMLR), 2023. [Tweet 1] [Tweet 2] [TechCrunch] [StarCoderBase] [StarCoder] [Github]

BigCode Team, “SantaCoder: don't reach for the stars!”, Deep Learning For Code workshop (DL4C) @ ICLR, 2023. [Best Paper Award] [Tweet 1] [Tweet 2] [SantaCoder]

卓 ㄓㄨㄛˊ越ㄩㄝˋ ?

Recent Publications

卓ㄓㄨㄛˊ越ㄩㄝˋ