Zory Zhang
Email: [email protected] Website: zoryzhang.notion.site
Education
University of Illinois, Urbana-Champaign (UIUC) Aug 2023 - May 2025 (Expected)
- Pursuing Bachelor's degree of Science in Mathematics and Computer Science, CGA=3.96/4.00
Hong Kong University of Science and Technology (HKUST) Sep 2021 - July 2023
- Former undergraduate majoring in Computer Science and Mathematics
![Me @Hong Kong West Kowloon Cultural District, 2022 Jan](https://s3-us-west-2.amazonaws.com/secure.notion-static.com/a0225eee-d607-471c-a37b-d69447ed34e4/Untitled.jpeg)
Me @Hong Kong West Kowloon Cultural District, 2022 Jan
Research Interest
I attempt to computationally characterize and model human learning and reasoning by asking what it takes to build (reverse engineering) a human-like (artificial) cognitive system: A system that can learn sample-efficiently, incrementally, and continually (without catastrophic forgetting), as well as flexibly generalize during performance.
I believe that reasoning is not only a performance task but also helps learning, i.e. Reason2Learn. Given prior belief in nearby domains, humans can learn so much from so few exposure to a novel domain (e.g. ARC). This mission bridges artificial intelligence and computational cognitive science, producing a flexible, explainable and persuasive system as a byproduct. My toolkit includes hierarchical concept formation (COBWEB), schema induction (LISA) and explanation-based learning (EBL) for human-like learning, analogical reasoning and “neuro-symbolic” architecture for human-like inference, as well as employing formal mathematics theorem proving as a playground to study cognition.
Publications
Under review:
- Zory Zhang, Ziyu Zhou, Haoran Liu, Jiaxuan You. (2024). Integrate Analogical Abduction, Connectionism and Explanation-based Learning in a Theorem Prover.
Accepted, in press:
- Dylan Zhang, Curt Tigges, Zory Zhang, Stella Biderman, Maxim Raginsky, and Talia Ringer (2024). Transformer-Based Models Are Not Yet Perfect At Learning to Emulate Structural Recursion. Transactions on Machine Learning Research. Arxiv.
Experience
- May 2024 - present: research intern
- Feb 2024 - present: research project leader
- Supervised by Jiaxuan You@UIUC
- Integrate Analogical Abduction, Connectionism and Explanation-based Learning in a Theorem Prover: idea formulation, 70%+ of coding, experiment monitoring and anlysis, visualization, and most of paper writing.
- Oct 2023 - Jan 2024: research assistant
- Supervised by Talia Ringer@UIUC and working with Dylan Zhang@UIUC
- Transformer-Based Models Are Not Yet Perfect At Learning to Emulate Structural Recursion**:** Design, impliment, and run three complementary experiments (toy transformer training distribution, large pre-trained language model finetuning, and GPT-4 prompting); result organization, analysis, and visualization.
- Mar 2023 - Aug 2023: research assistant
- Supervised by Yangqiu Song@HKUST and working with Zihao Wang@HKUST
- A Knowledge graph reasoning via ML project: Provided careful evaluation of the theoretical framework, proposed one of the two filtering modules to overcome training difficulty, and conducted massive coding and experiments.
Related Courses
- Philosophy/psychology courses:
- UIUC: Reasoning and Thinking[A] (focuses on reasoning about scientific enquiry), Logic and Thinking[A](focuses on general training on analyzing and writing logical arguments), Neural Network Modeling[Sit-in] (to study priciples in computational modeling).
- Intensive Deep Learning courses and projects:
- HKUST: Deep Learning in Computer Vision [A], Machine Learning with Structured Data [A].
- Various rigorous mathematics courses:
- UIUC: (graduate) Real Analysis[A], Graph Theory[A].
- HKUST: Honors Calculus I[A+] & II, Multivariable Calculus[A], Honors Mathematical Analysis[A], Honors Discrete Mathematical Tools for Computer Science [A+], Honors in Linear and Abstract Algebra I [A], and Honors Probability [A-].
- Broad computer science courses and projects:
- Typical computer science curriculum.
- HKUST: Advanced Algorithms [A]: a graduate algorithm course
- UC Berkeley: ****Operating Systems[A] and Machine Architecture [A-] in one single summer
Extracurricular Activities
Academic
- Member of HKUST Programming Team Sep 2021-Aug 2022
- In International Collegiate Programming Contest(ICPC) competitions, one must solve problems effectively with another two teammates in a limited time. Collaborative work highlights interpersonal communication, cooperation, demonstration, illustration, and time management skills.
- Our team won 52-rd place in the ICPC Asia Regional Contest Jinan Site 2021, with a Silver medal; Our team won runner-up in the Tsinghua-HKUST Programming Contest 2022
- Personal achievement: Rating 2313 at codeforeces
- Undergraduate Teaching Assistant of COMP2211 Exploring Artificial Intelligence Sep-Dec 2022
Leadership & Communication skills related
- Executive Committee of HKUST Drama Society 2022-2023
- Gained hands-on experience in organizing and leading drama events