Juseon-Do

Research Insterest

• Natural Language Processing (NLP)

• Summarization

• Retrieval Augmented-Summarization (RAG)

• Prompt Engineering

• Agent Modeling

• MLOps

Education

• Chungnam National University - BS

March, 2019  ~  Feb, 2026

Career

• GILAB(Chungnam National University) - Undergraduate AI Researcher

November, 2023  ~  Current

• AIST(National Institute of Advanced Industrial Science and Technology) - Intern

June, 2024  ~  Aug, 2024

Publications

1. InstructCMP: Length Control in Sentence Compression through Instruction-based Large Language Models(ACL 2024 Findings)

Juseon-Do, Jingun Kwon, Hidetaka Kamigaito, Manabu Okumura

2. Considering Length Diversity in Retrieval-Augmented Summarization(NAACL 2025 Findings) 

Juseon-Do, Jaesung Hwang, Jingun Kwon, Hidetaka Kamigaito, Manabu Okumura

3. AAAI2026: Under Reviewing

Juseon-Do, Sungwoo Han, Jingun Kwon, Hidetaka Kamigaito, Manabu Okumura

Projects

1. Factual Summarization (Generation Pipeline Modification)

The method generates multiple candidate sentences at each step and selects the most factually consistent ones by comparing them with the source document. Unlike traditional beam search, pruning is applied at the sentence level, which improves factuality. Implementing this in batch mode required a solid understanding of padding, past key values, and positional encoding. Github

2. Korean SAT App (Agent Modeling & MLOps)

Developed an application that transforms English passages into diverse question types for the Korean SAT. Designed multiple agents with distinct personas to handle question generation, feedback, and refinement, resulting in highly accurate and high-quality exam questions.

3. Bridge Image and Text (Prompt Engineering)

Enhanced multi-modal model performance by improving the understanding of relationships between images and text.

Academic Service

1. Reviewer for AAAI2026

Donation

1. Hair for children with cacner(Aug. 29. 2025)