About Me

I am a second-year Ph.D. student in the Optimization & Machine Learning (OptiML) Laboratory at Kim Jaechul Graduate School of AI, Korea Advanced Institute of Science and Technology (KAIST AI), where I am fortunate to be advised by Prof. Chulhee Yun. Previously, I completed my M.Sc. in AI at KAIST and B.E. in Cyber Security (minor in Mathematics) at Korea University.

Research Interests

I study the theoretical foundations of modern deep learning and design algorithms grounded in rigorous analysis. My work [C1, C4] establishes lower bounds on convergence rates for without-replacement SGD. Another line of work [C2, C3] develops and theoretically justifies an algorithm that strengthens length generalization in Transformers. Current projects investigate speculative decoding, where I am working on designing sampling strategies to improve inference efficiency given a tree-structured draft. I am also studying formal constructions showing how Transformers can perform in-context inference in hidden Markov models.

News

• [May. ‘25] Our paper on Convergence Rate of Shuffling SGD was accepted to ICML 2025.
• [Nov. ‘24] I was selected as one of the Top Reviewers at NeurIPS 2024 (Top 1,303/15,160 = 8.6% of reviewers).
• [Sep. ‘24] Our paper on Length Generalization of Arithmetic Transformers was accepted to NeurIPS 2024.
• [Jun. ‘24] Our earlier work on Length Generalization of Arithmetic Transformers was accepted to ICML 2024 Workshop on Long-Context Foundation Models.
• [Feb. ‘24] I completed my master’s degree and began my Ph.D. journey.
• [Jun. ‘23] Our paper on Convergence Rate of Shuffling SGD was accepted to ICML 2023 with an Oral Presentation (Top 155/6,538 = 2.37% of papers).
• [Feb. ‘22] I started my master’s degree at the OptiML Lab at KAIST AI.

Academic Services

Conference Reviewer

• NeurIPS 2023, 2024, 2025
• ICLR 2025, 2026
• ICML 2025

Journal Reviewer

• TMLR 2025
• JOTA 2025

Education

• Ph.D. in Artificial Intelligence, KAIST (Mar. 2024 – Present)
• M.Sc. in Artificial Intelligence, KAIST (Mar. 2022 – Feb. 2024)
• B.E. in Cyber Security, Korea University (Mar. 2018 – Feb. 2022)
  • Minor in Mathematics

Contact

• CV: PDF
• Email: chajaeyoung at kaist dot ac dot kr