Jungseob Lee hiking in the mountains Jungseob Lee
NLP Researcher

Jungseob Lee (이정섭)

Graduate Student, Korea University

I am a graduate student at Korea University, advised by Prof. Heuiseok Lim. My research focuses on Natural Language Processing, with particular interests in large language models, natural language processing, agentic LLM, machine translation, information retrieval, and model merging.

Large Language Models Natural Language Processing Agentic LLM Machine Translation Information Retrieval Model Merging

News

2026
IEEE BIBM 2026 Our paper "Structured Clinical Memory for Safety-Consistent Diagnosis and Recommendation Generation in Long Medical Dialogues" has been accepted at IEEE BIBM 2026 (Short Paper).
2026
Findings of AACL-IJCNLP 2026 Our paper "Skin-Deep: A Geometric Diagnostic for Alignment Fragility in Large Language Model Representations" has been accepted at Findings of AACL-IJCNLP 2026 (co-first author).
2026
EMNLP 2026 Four papers (2 Main, 2 Findings) have been accepted at EMNLP 2026.
2026
EMNLP 2026 Our paper "The Hallucination Signal Is a Mean Shift: Why Simple Probes Suffice" has been accepted at EMNLP 2026 Main.
2026
EMNLP 2026 Our paper "Language Chain in Alignment: Cross-Lingual Ranking Preference Optimization" has been accepted at EMNLP 2026 Main.
2026
Findings of EMNLP 2026 Our paper "DART: Draft-Agreement Routing for Training-Free Adaptive Thinking Budgets in Hybrid Reasoning Models" has been accepted at Findings of EMNLP 2026.
2026
Findings of EMNLP 2026 Our paper "ThinkFuse: Trajectory-Aware Test-Time Fusion for Small Reasoning Models" has been accepted at Findings of EMNLP 2026 (co-first author).
2026
ACL 2026 Received the Outstanding Reviewer Award at ACL 2026 (the 64th Annual Meeting of the Association for Computational Linguistics), San Diego.
2026
KBS 2026 Our paper "Towards enhancing Natural Language Inference for identifying erroneous outputs in data-to-text generation" has been accepted at Knowledge-Based Systems (KBS).
2026
ICLR 2026 Our paper "Improving Semantic Proximity in Information Retrieval through Cross-Lingual Alignment" has been accepted at ICLR 2026 (Poster).

Preprints

Publications

Projects

Led by NC AI (NCSOFT) with a consortium of five Korean enterprises · First release: Dec 30, 2025
Post-training Project Manager — Korea University Representative
  • Managed the full post-training pipeline including Supervised Fine-Tuning (SFT), Reinforcement Learning, and dataset curation as Project Manager representing Korea University
  • VAETKI achieves state-of-the-art performance on Korean language and cultural comprehension (CLIcK), advanced reasoning (KoBALT), professional knowledge (KMMLU-Pro), and instruction following (IFEval)
VAETKI benchmark performance comparison

Awards & Honors

2026

Outstanding Reviewer Award

The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026), San Diego

Recognized by the ACL program committee for review quality and service to the conference.

2025

Best Paper Award (Top 1)

Annual Conference on Human and Cognitive Language Technology (HCLT 2025)

For "KULLM-RAG: A RAG-Specialized Large Language Model with Intelligent Self-Verification" (first author)
Korea University · Soongsil University · Human-Inspired AI Research

2023

Minister of Science and ICT Award

ICT Challenge 2023 — Ministry of Science and ICT (organized by IITP)

Team Myunghoon Kang, Jungseob Lee, Seungyoon Lee, for "KoCheckGPT" — Korean-specialized detection of LLM-generated documents for enterprise security. Selected from 79 teams (263 participants) across 28 universities nationwide.

Education

2021 – Present

M.S. / Ph.D. in Computer Science

Korea University

NLP Lab — Advisor: Prof. Heuiseok Lim
Research: Large Language Models, Natural Language Processing, Agentic LLM, Machine Translation, Information Retrieval, Model Merging