Heegyu Kim

AI Scientist @ LG AI Research

Master's Degree in Artificial Intelligence @ Ajou University

Email: heekue83@gmail.com

Research Interests

LLM Agent for Materials Science and Database (Text-to-SQL)
Safe and Reliable Language Models
LLM-as-a-Judge

Education

Ajou University, Suwon, South Korea

M.S. in Artificial Intelligence, 2023 - 2025 (Advisor: Hyunsouk Cho)
B.S. in Software Engineering, 2016 - 2023

Work Experience

LG AI Research, Seoul, South Korea

AI Scientist (2025.7 - Present)

Research Engineer Internship (2025.2 - 2025.6)
Undergraduate Internship, 2022.7 - 2022.12 (Mentor: Seonghwan Kim)

Ably Corporation, Seoul, South Korea

Android Developer, 2020

BitByte Corporation, Seoul, South Korea

Android Developer, 2017-2019

Publications

Kim, Heegyu, et al. "Towards Fully-Automated Materials Discovery via Large-Scale Synthesis Dataset and Expert-Level LLM-as-a-Judge." arXiv preprint arXiv:2502.16457 (2025) - Accepted into CIKM 2025

Kim, Heegyu, et al. "FLEX: Expert-level False-Less EXecution Metric for Reliable Text-to-SQL Benchmark." arXiv preprint arXiv:2409.19014 (2024) - Accepted into NAACL 2025 (Oral).

Heegyu Kim, Sehyun Yuk, and Hyunsouk Cho. "Break the Breakout: Reinventing LM Defense Against Jailbreak Attacks with Self-Refinement." arXiv preprint arXiv:2402.15180 (2024) - Accepted into NAACL 2025 Workshop TrustNLP.

Heegyu Kim and Hyunsouk Cho. 2023. "GTA: Gated Toxicity Avoidance for LM Performance Preservation." In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 14747–14763, Singapore. Association for Computational Linguistics.

Awards

Bronze, Opensource Software Developer Competition (2023)

1st, Shinhan Bigdata Hackathon (2022)

Open Source Contributions

Lectures

Sunrin Internet High School, Seoul, South Korea

Instructor, Special Lecture on Artificial Intelligence
Winter 2022-2023, Winter 2023-2024, Winter 2024-2025