About Me

I am a Ph.D. student at KAIST, advised by Prof. Jong Chul Ye. I study how to make language models reason, retrieve evidence, and respond to rewards more effectively.

My research focuses on language model reasoning, retrieval-augmented generation, reinforcement learning, and reward-guided model behavior, with applications in general reasoning and clinical AI.

News

  • [Sep. 2026] LEGO-OPD released on arXiv.
  • [Sep. 2026] Dementia-R1 selected for an oral presentation at EMNLP 2026.
  • [Apr. 2026] Universal Reasoner accepted to ICML 2026.
  • [Mar. 2026] Hypothesis-Conditioned Query Rewriting released on arXiv.
  • [Feb. 2026] Received the Silver Prize at the 32nd Samsung Humantech Paper Award.
  • [Feb. 2026] Received the Silver Prize in the IPIU 2026 Best Paper Award.
  • [Jan. 2026] Dementia-R1 released on arXiv.

Publications

* equal contribution
ICML 2026

Universal Reasoner: A Single, Composable Plug-and-Play Reasoner for Frozen LLMs

Jaemin Kim*, Hangeol Chang*, Hyunmin Hwang*, Choonghan Kim, Jong Chul Ye

A lightweight reward-trained module that adds composable reasoning skills to frozen language models.

arXiv
Under review · 2026

LEGO-OPD: Factorized Teacher Composition for Multimodal On-Policy Distillation

Jaeyun Shin*, Hangeol Chang*, Jong Chul Ye

Factorized teacher composition that improves visual grounding while preserving text-only reasoning in multimodal on-policy distillation.

arXiv
Under review · 2026

Hypothesis-Conditioned Query Rewriting for Decision-Useful Retrieval

Hangeol Chang, Changsun Lee, Seungjoon Rho, Junho Yeo, Jong Chul Ye

A training-free retrieval planner that searches for evidence useful to a decision, rather than broadly relevant context.

arXiv
EMNLP 2026 · Oral

Dementia-R1: Reinforced Pretraining and Reasoning from Unstructured Clinical Notes

Choonghan Kim*, Hyunmin Hwang*, Hangeol Chang*, Jaemin Kim*, Jinse Park, Jae-Sung Lim, Jong Chul Ye

Longitudinal clinical reasoning with verifiable intermediate rewards for real-world dementia prognosis.

arXiv
ICML Workshop 2026

AgentPSO: Evolving Agent Reasoning Skill via Multi-agent Particle Swarm Optimization

Hyunmin Hwang*, Jaemin Kim*, Choonghan Kim, Hangeol Chang, Jong Chul Ye

A particle-swarm-inspired framework for evolving reusable reasoning skills across a population of agents.

arXiv
IPIU 2026

Ground-A-Score: Scaling Up the Score Distillation for Multi-Attribute Editing

Hangeol Chang*, Jinho Chang*, Jong Chul Ye

Grounded score distillation for precise image edits with multiple attributes and spatial constraints.

arXiv

Honors

2026
Silver Prize · 32nd Samsung Humantech Paper Award

Universal Reasoner · 2nd Prize in Signal Processing

2026
Silver Prize · Best Paper Award at IPIU 2026

Ground-A-Score · Top 2% paper

Education

2024—
KAIST

Ph.D. in Artificial Intelligence · Kim Jaechul Graduate School of AI

LLM reasoning, retrieval-augmented generation & reinforcement learning · Advisor: Prof. Jong Chul Ye

2022—24
KAIST

M.S. in Artificial Intelligence · Kim Jaechul Graduate School of AI

Diffusion models & LLM-guided image editing · Advisor: Prof. Jong Chul Ye

2021—22
Yonsei University

Graduate Studies in Physics · Department of Physics

Solid-state physics & ARPES · M.S.–Ph.D. integrated program, three semesters

2018—21
Yonsei University

B.S. in Integrated Technology · School of Integrated Technology

Full scholarship · Early graduation · GPA 3.95/4.3

Research Projects

2025
Clinical Decision Support with RAG

On-site research project with NAVER · Physician-support pipeline

2024—
LLM-based Dementia Data Processing & Clinical Support

Government-funded research project