Hanseul Cho

Hanseul Cho

Student Researcher in Google NYC. PhD Student at OptiML Lab in KAIST AI. Working on ML/DL/Opt+Theory and Language Models.

Hanseul Cho (์กฐํ•œ์Šฌ) ๐Ÿ‘‹

I am a Ph.D. student at Kim Jaechul Graduate School of AI, Korea Advanced Institute of Science and Technology (KAIST AI) (08/2023โ€“08/2027 (expected)). I am fortunate to be advised by Prof. Chulhee โ€œCharlieโ€ Yun of the Optimization & Machine Learning (OptiML) Laboratory, KAIST AI. Previously, I completed my M.Sc. (in AI, 02/2022โ€“08/2023) and B.Sc. (in Math, minor in CS, Summa Cum Laude) at KAIST.

Currently, I am a Student Researcher at Google in New York City๐Ÿ‡บ๐Ÿ‡ธ (05/05/2025โ€“07/25/2025), working for Srinadh Bhojanapalli. Also, Iโ€™ll be in NYC from 05/02/2025 to 08/21/2025, exploiting the grace periods before & after my J-1 visa program.

Please donโ€™t hesitate to reach out for questions, discussions, and collaborations! ๐Ÿค—

๐Ÿ“‹ Curriculum Vitae (CV): [PDF] | [Overleaf-ReadOnly]
๐Ÿ“ง Primary E-mail: jhs4015 at kaist dot ac dot kr
๐Ÿ“ง Googler E-mail: {firstname}{lastname} at google dot com

๐Ÿ”ฌ Research Interests ๐Ÿ”ญ

My primary research interests lie in optimization, machine learning (ML), and deep learning (DL), especially focusing on both mathematical/theoretical analysis and empirical improvements (usually based on theoretical understanding).

During my journey to a Ph.D.๐Ÿ‘จ๐Ÿปโ€๐ŸŽ“, my ultimate research goal is to rigorously understand and practically overcome the following three critical challenges in ML/DL (see my Thesis Proposal๐Ÿ”—๐Ÿ“ฐ if interested):

  • [Generalizability]

    Out-of-distribution generalization of (large) language models (e.g., length generalization and compositional generalization of Transformers)

  • [Adaptability]

    Training adaptable models under evolving environments (e.g., continual learning, maintaining the plasticity of neural networks, sample-efficient reinforcement learning)

  • [Multifacetedness]

    Learning with multiple (possibly conflicting and/or orthogonal) goals (e.g., minimax optimization, biโ€‘level optimization, fairness in ML)

โ€ผ๏ธNewsโ€ผ๏ธ

  • ๐Ÿ—ž๏ธ [May '25] (NEW) I visit NYC๐Ÿ‡บ๐Ÿ‡ธ from 2025-05-02 to 2025-08-21 (see the item below). Let's grab a coffee and have a chat if you are in NYC!
  • ๐Ÿ—ž๏ธ [Feb. '25] I'll work as a Student Researcher at Google in New York City๐Ÿ‡บ๐Ÿ‡ธ! (05/05/2025–07/25/2025, Host: Srinadh Bhojanapalli)
  • ๐Ÿ—ž๏ธ [Jan. '25] Invited as a reviewer of Transactions on Machine Learning Research (TMLR).
  • ๐Ÿ—ž๏ธ [Jan. '25] Two papers got accepted to ICLR 2025! ๐ŸŽ‰ One is the sequel of our Position Coupling paper; another is about a theoretical analysis of continual learning algorithm. See you in Singapore๐Ÿ‡ธ๐Ÿ‡ฌ!
  • ๐Ÿ—ž๏ธ [Nov. '24] An early version of our paper on theoretical analysis of continual learning is accepted to JKAIA 2024 and won the Best Paper Award (top 3 papers)! ๐ŸŽ‰
  • ๐Ÿ—ž๏ธ [Nov. '24] I'm selected as one of the Top Reviewers (top 8.6%: 1,304 of 15,160 reviewers) at NeurIPS 2024! (+ Free registration! ๐Ÿ˜Ž)
  • ๐Ÿ—ž๏ธ [Sep. '24] Two papers got accepted to NeurIPS 2024! ๐ŸŽ‰ One is about length generalization of arithmetic Transfomers, and another is about mitigating loss of plasticity in incremental neural net training. See you in Vancouver, Canada๐Ÿ‡จ๐Ÿ‡ฆ!
  • ๐Ÿ—ž๏ธ [Jun. '24] An early version of our paper on length generalization of Transformers got accepted to the ICML 2024 Workshop on Long-Context Foundation Models!
  • ๐Ÿ—ž๏ธ [May '24] A paper got accepted to ICML 2024 as a spotlight paper (top 3.5% among all submissions)! ๐ŸŽ‰ We show global convergence of Alt-GDA (which is strictly faster than Sim-GDA) and propose an enhanced algorithm called Alex-GDA for minimax optimization. See you in Vienna, Austria๐Ÿ‡ฆ๐Ÿ‡น!
  • ๐Ÿ—ž๏ธ [Sep. '23] Two papers are accepted to NeurIPS 2023! ๐ŸŽ‰ One is about Fair Streaming PCA and another is about enhancing plasticity in RL. See you in New Orleans, USA๐Ÿ‡บ๐Ÿ‡ธ!
  • ๐Ÿ—ž๏ธ [Jan. '23] Our paper about shuffling-based stochastic gradient descent-ascent got accepted to ICLR 2023!
  • ๐Ÿ—ž๏ธ [Nov. '22] An early version of our paper about shuffling-based stochastic gradient descent-ascent is accepted to 2022 Korea AI Association + NAVER Autumnal Joint Conference (JKAIA 2022) and selected as the NAVER Outstanding Theory Paper (top 3 papers)!
  • ๐Ÿ—ž๏ธ [Oct. '22] I am happy to announce that our very first preprint is now on arXiv! It is about convergence analysis of shuffling-based stochastic gradient descent-ascent.
  • ๐Ÿ—ž๏ธ [Feb. '22] Now I am part of OptiML Lab of KAIST AI.