Jinho Park

About

Hello👋🏻, I'm a M.S. student in Language & Knowlege Lab at KAIST Kim Jaechul Graduate School of AI. I am grateful to be advised by Minjoon Seo. Before my Master's, I earned a B.S. in Computer Science at KAIST with a minor in Electrical Engineering in 2024.

I am passionate about analyzing and applying language models within the field of Natural Language Processing (NLP). My research focuses on exploring how language models acquire and utilize knowledge during their training, specifically through the areas of Knowledge Acquisition of LM and Mechanistic Interpretability. I aim to build reliable language models based on a deeper, more robust understanding of their internal process. Additionally, I have an interest in advancing high-performing large language models (LLMs) in areas such as Multilingual and Reasoning, as well as exploring research in State Space Models (SSM).

For details, refer to my Curriculum Vitae (CV).

Publications

Most recent publications on Google Scholar.
^‡ indicates equal contribution.

How Do Large Language Models Acquire Factual Knowledge During Pretraining?

Hoyeon Chang, Jinho Park, Seonghyeon Ye, Sohee Yang, Youngkyung Seo, Du-Seong Chang, Minjoon Seo

NeurIPS, 2024

Paper Code Bib

 
                  
 @inproceedings{chang2024large,
 title={{How do large language models acquire factual knowledge during pretraining?}},
 author={Chang, Hoyeon and Park, Jinho and Ye, Seonghyeon and Yang, Sohee and Seo, Youngkyung and Chang, Du-Seong and Seo, Minjoon},
 booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems}
 }

Cognitive map for language models: Optimal planning via verbally representing the world model

Doyoung Kim, Jongwon Lee, Jinho Park, Minjoon Seo

CoRR 2024

Paper Code Bib

 
                  
 @article{kim2024cognitive,
 title={Cognitive Map for Language Models: Optimal Planning via Verbally Representing the World Model},
 author={Kim, Doyoung and Lee, Jongwon and Park, Jinho and Seo, Minjoon},
 journal={CoRR},
 year={2024}
 }

Projects

Selected
All

Shoulder of Giants
Reimplement the paper (3D AI)

HOMEY
Mobile App optimizing home environment

Training Efficient Boosting Algorithm
Predict fake import declarations using LightGBM

SPARCS Taxi
Development on-campus taxi carpool App using React.js and Javascript

Developing an in-house analytics page
Event log analysis page using Vue.js and Spring

Kaggle Competition
Detect distracted driver using Deep Learning

Creating Song Lyrics System
Using Char-RNN model (Deep Learning)