I’m a PhD Student at the Linguistic Department at Georgetown University. I am most interested in natural language understanding and how LLMs and computational systems encode linguistic structure to perform complex NLU tasks. I am also interested in language modeling beyond text, LLMs as models of cognition, and treebanking.

Before coming to Georgetown, I was NLP engineer at NCSOFT, where I worked on NLU and information extraction.

Outside of classes and assitantships, I like to participate in intramural sports as both player and coach, attend free food events around campus, and catch Thursday afternoon flights to Toronto.

For Summer 2025, I will be joining the Lee Language Lab (University of Toronto, Ontario Tech University) as a visiting researcher.

You can reach me at jm3743@georgetown.edu.

Publications

Junghyun Min, Minho Lee, Woochul Lee, Yeonsoo Lee. Punctuation restoration improves structure understanding without supervision. RepL4NLP at NAACL 2025.

Junghyun Min, R. Thomas McCoy, Dipanjan Das, Emily Pitler, and Tal Linzen. Syntactic data augmentation increases robustness to inference heuristics. ACL 2020.

R. Thomas McCoy, Junghyun Min, and Tal Linzen. BERTs of a feather do not generalize together: Large variability in generalization across models with similar test set performance. BlackboxNLP at EMNLP 2020.

Education and work experience

Ph.D. student, Computational Linguistics, Georgetown University2024 -
NLP Engineer, NCSOFT2021 - 2024
M.A. Cognitive Science, Johns Hopkins University. Advised by Tal Linzen2019 - 2020
Data Analyst, Harford Community College2018 - 2019
B.S. Physics, B.A. Mathematics, Johns Hopkins University2014 - 2017