Jeonghwan Choi

Jeonghwan Choi

Ph.D. Student

Industrial & Systems Engineering
KAIST · Data Intelligence System Lab

Research Interests

  • Retrieval-Augmented Generation
  • LLM Evaluation & Reliability
  • Human-AI Alignment
  • LLM Agent Simulation
  • Physical AI

About Me

I am a Ph.D. student at the Data Intelligence System Lab (DISL) at KAIST, where I am advised by Prof. Hwanjun Song. I first joined the lab for my M.S. and then continued into the integrated Ph.D. program, after studying Economics with a double major in Statistics at Yonsei University.

My work centers on Retrieval-Augmented Generation, where I build frameworks that keep generation faithful to retrieved evidence and design evaluation methods that reveal when they fail. The same interest carries into human-AI alignment and LLM agent simulation, and more recently into Physical AI, where I am exploring world model-based robot learning and sim-to-real transfer.

Open to collaboration. I'm always happy to discuss RAG, LLM evaluation, agent simulation, and world models for Physical AI. Feel free to reach out by email.

Research Overview

Robust RAG Frameworks

Retrieval and generation that stays faithful to source evidence, through query expansion and extraction-generation alignment.

Evaluation & Reliability

Benchmarks and automatic evaluators that expose where RAG systems break, from query-agnostic metrics to multi-agent relevance annotation.

LLM Agent Simulation

Language models that simulate multi-party behavior, including seller-buyer retail dynamics and multimodal conflict mediation.

Physical AI

World model-based robot learning and sim-to-real transfer, extending agent reasoning to systems that act in the physical world.

Latest News

2026.08
Selected for the Doctoral Student Research Grant from the National Research Foundation of Korea as Principal Investigator, KRW 50,000,000 over two years for work on self-evolving Physical AI.
2026.08
Towards Robust RAG via Reliability-Aware Adaptive Evidence Selection accepted to Findings of EMNLP 2026.
2026.07
Three papers accepted to COLM 2026. RetailSim · Q-CARE · RAQE
2026.05
Distilling Long-CoT Reasoning accepted to Findings of ACL 2026. [paper]
2026.01
Completing Missing Annotation accepted to ICLR 2026, my first ICLR paper. [OpenReview]
2025.12
Best Paper Award at Korea Software Congress 2025.
2025.10
Aligning Extraction and Generation for Robust RAG accepted to WSDM 2026. [arXiv]
2025.06
Word2Passage accepted to Findings of ACL 2025. [paper]

Education

Ph.D. Program, Industrial & Systems Engineering

Feb 2025 - Present · Daejeon, South Korea

KAIST · Advisor: Prof. Hwanjun Song · Integrated from M.S.

M.S., Graduate School of Data Science

Feb 2024 - Feb 2025 · Daejeon, South Korea

KAIST · Advisor: Prof. Hwanjun Song · GPA 4.22 / 4.3

B.A. Economics, Statistics Double Major

Until Feb 2024 · Seoul, South Korea

Yonsei University · GPA 4.21 / 4.3

Publications

* denotes equal contribution

Semantic Scholar
Under Review(2)
  • Toward Agentic Retrieval-Augmented Generation for Enterprise Retail Workflows over Heterogeneous Data Sources First Author
    Jeonghwan Choi, Jibin Hwang, HaYoon Park, Hyeonjae Cheon, Hwanjun Song
    Under Review · 2026
  • SoCRATES: Towards Reliable Automated Evaluation of Proactive LLM Mediation across Domains and Socio-cognitive Variations
    Taewon Yun, Hyeonseong Park, Jeonghwan Choi, HaYoon Park, Yeeun Choi, Hwanjun Song
    Under Review · 2026
    arXiv
2026(7)
  • Towards Robust Retrieval-Augmented Generation via Reliability-Aware Adaptive Evidence Selection
    Minkyu Che, Jihwan Oh, Jeonghwan Choi, Jeesu Jung, Hwanjun Song
    Findings of EMNLP 2026
  • What Makes a Sale? Simulating End-to-End Seller-Buyer Retail Dynamics with LLM Agents First Author
    Jeonghwan Choi, Jibin Hwang, Gyeonghun Sun, Minjeong Ban, Taewon Yun, Hwanjun Song
    COLM 2026
    arXiv
  • Towards Query-Agnostic RAG Evaluation via Query Coverage and Claim Verifiability First Author
    Jeonghwan Choi, Taewon Yun, Minjeong Ban, Gyeonghun Sun, Jae-gil Lee, Hwanjun Song
    COLM 2026
    arXiv code
  • RAQE: Reranker-Aligned Query Expansion via Label-Free Group-Relative Policy Optimization
    Gyeonghun Sun, Jeonghwan Choi, Hwanjun Song
    COLM 2026
  • Completing Missing Annotation: Multi-Agent Debate for Accurate and Scalable Relevant Assessment for IR Benchmarks First Author
    Jeonghwan Choi*, Minjeong Ban*, Hyangsuk Min*, Nicole Hee-yeon Kim, Minseok Kim, Jae-gil Lee, Hwanjun Song
    ICLR 2026
    PDF
  • Distilling Long-CoT Reasoning through Collaborative Step-wise Multi-Teacher Decoding
    Taewon Yun, Jisu Shin, Jeonghwan Choi, Seunghwan Bang, Hwanjun Song
    Findings of ACL 2026
    PDF
  • Aligning Extraction and Generation for Robust Retrieval-Augmented Generation
    Hwanjun Song, Jeonghwan Choi, Minseok Kim
    WSDM 2026
    PDF
2025(3)
  • BRIDGE: Toward Faithful RAG Benchmarking via Retrieval-Generation Alignment First Author
    Jeonghwan Choi*, Minjeong Ban*, Hyangsuk Min*, Hee-yeon Kim, Seunghwan Bang, Hwanjun Song
    RDGENAI@CIKM 2025
  • Word2Passage: Word-level Importance Re-weighting for Query Expansion First Author
    Jeonghwan Choi, Minjeong Ban, Minseok Kim, Hwanjun Song
    Findings of ACL 2025
    PDF
  • Learning to Verify Summary Facts with Fine-Grained LLM Feedback
    Jihwan Oh, Jeonghwan Choi, Nicole Hee-Yeon Kim, Taewon Yun, Hwanjun Song
    COLING 2025
    PDF
Domestic(9)
  • Simulating Seller-Buyer Retail Interactions and Sales Strategies with LLM Agents First Author
    Jeonghwan Choi, Jibin Hwang, Gyeonghun Sun, Minjeong Ban, Taewon Yun, Hwanjun Song
    KCC 2026
  • Toward Low-Cost Query Expansion: Label-Free Reward Design via Reranker Scores
    Gyeonghun Sun, Jeonghwan Choi, Sundong Kim, Hwanjun Song
    KCC 2026
  • A Query Decompositional Framework for Fine-grained RAG Assessment First Author
    Jeonghwan Choi, Minjeong Ban, Taewon Yun, Jihwan Oh, Yuho Lee, Hwanjun Song
    KTCP 2026 · JournalExtended version of KSC 2025 Best Paper
  • A Query Decompositional Framework for Fine-grained RAG Assessment Best Paper
    Jeonghwan Choi, Minjeong Ban, Nicole Hee-Yeon Kim, Taewon Yun, Yuho Lee, Hwanjun Song
    KSC 2025
  • A Benchmark Dataset for Retrieval-Augmented Generation Considering Performance Alignment between Retrieval and Generation Best Presentation
    Jeonghwan Choi*, Minjeong Ban*, Hyangsuk Min*, Taewon Yun, Jihwan Oh, Hwanjun Song
    KCC 2025
  • Robust Dataset Condensation via Semi-Supervised Learning
    Hee-yeon Kim, Jeonghwan Choi, Yuho Lee, Hwanjun Song
    KCC 2025
  • Improving Language Model Quality through LLM-based Fine-Grained Hallucinated Summary Generation
    Jihwan Oh, Jeonghwan Choi, Nicole Hee-Yeon Kim, Taewon Yun, Hwanjun Song
    KTCP 2025 · Journal
  • Training Summary Evaluation Model with Fine-Grained LLM Feedback Best Presentation
    Jihwan Oh, Jeonghwan Choi, Taewon Yun, Hyangsuk Min, Hwanjun Song
    KSC 2024
  • Improving the Text Summary Quality Through Understanding the Hallucination Level of Summarization Using Large Language Models Best Presentation
    Jihwan Oh, Jeonghwan Choi, Nicole Hee-Yeon Kim, Hwanjun Song
    KCC 2024

Patents

Cost-Efficient and Accurate Debate-Based Relevance Assessment System with Multi-Agents

Apr 2026 · KIPO · Application No. 10-2026-0066617

Inventors: Jeonghwan Choi (20%), Minjeong Ban (20%), Hyangsuk Min (20%), Hee-yeon Kim (10%), Jae-gil Lee (10%), Hwanjun Song (20%)

Method and Apparatus for Generating Answers Using Extended Query-Based Retrieval Utilizing Multi-Level Importance Reweighting

Sep 2025 · KIPO · Application No. 10-2025-0127517

Inventors: Jeonghwan Choi (40%), Minjeong Ban (20%), Hwanjun Song (40%)

Honors & Awards

2026

Doctoral Student Research Grant

National Research Foundation of Korea (NRF), Ministry of Education · Principal Investigator

2025

Best Paper

Korea Software Congress (KSC)

2025

Best Presentation

Korea Computer Congress (KCC)

2024

Best Presentation

Korea Software Congress (KSC)

2024

Best Presentation

Korea Computer Congress (KCC)

2023

National Scholarship for Humanities Excellence

Korea Student Aid Foundation (KOSAF)

2023

Highest Honors (Top 1%)

Yonsei University

2023

Guwon Foundation Scholarship

Guwon Scholarship Foundation · Top 1 in Department

2023

High Honors (Top 3%)

Yonsei University

2022

High Honors (Top 3%)

Yonsei University

Academic Service, Teaching & Leadership

Program Committee Member: LASS 2026

2nd Workshop on LLM Agents for Social Simulation · Co-located with ACM CIKM 2026

Reviewer: DocInsights 2026

Workshop on Document Intelligence and Understanding · Co-located with EMNLP 2026

Teaching Assistant: Dynamic Programming and Reinforcement Learning (IE540)

KAIST · Spring 2025

President, Y-FoRM (Yonsei Forum of Risk Management)

Sep 2022 - Aug 2023 · AI-driven Financial Prediction Research Society

Led 30 members and delivered lectures on Statistics, Calculus, Linear Algebra, and Deep Learning.

Certifications

2020 - 2022

  • ADsP (Advanced Data Analytics Semi-Professional), Korea Data Agency, Mar 2022
  • FRM Part 1 (Financial Risk Manager), GARP, Jul 2021
  • Investment Asset Management Professional, Korea Financial Investment Association, Jun 2020

Skills

Programming: Python, Java, R  ·  Languages: English (Proficient), Korean (Native)

Project Experience

Self-Evolving Physical AI

Principal Investigator · National Research Foundation of Korea · Sep 2026 - Aug 2028

Developing agents that continually learn from real-world experience after deployment. Doctoral Student Research Grant, KRW 50,000,000 over two years.

Retrieval-Augmented Generation for Retail

Project Manager · GS Retail · Jan 2026 - Jun 2026

Developing a retail-specialized RAG framework with retail environment simulation.

Multimodal Agentic Platform for Conflict Mediation

Research Team Member · Ministry of Science and ICT (IITP) · Oct 2025 - Present

Developing a multimodal agentic AI for real-time conflict mediation using text, speech, and vision signals.

Korea-Canada Joint Research (KAIST-GIST-UBC)

Research Team Member · IITP · Sep 2024 - Present

Building value-aligned benchmark datasets for language, math, and medical domains.

Language Model Reliability Enhancement System

Research Team Member · National Research Foundation of Korea (NRF) · Apr 2024 - Present

Developing evaluation systems for AI-generated text and reducing hallucinations in LLMs.

Scientific Information RAG System

Project Manager · KISTI · Jun 2025 - Dec 2025

Optimized RAG pipeline for scientific domains, maximizing performance while reducing latency.

Korean Scientific RAG Framework Development

Project Manager · KISTI · Jun 2024 - Dec 2024

Developed and evaluated a RAG framework optimized for Korean scientific domains.

Contact

Email

hwani.choi@kaist.ac.kr