YongHwi Song

YongHwi Song

송용휘

B.S. Candidate, Chungbuk National University
School of Information and Communication Engineering · Cheongju, Korea

I am an undergraduate researcher at Chungbuk National University and a research intern at the Artificial Intelligence Laboratory. My research centers on world models and world-action models: how an agent can learn a compact model of its environment, plan inside it, and read actions out of it cheaply. I study where these models fail under efficiency constraints, such as planners exploiting the errors of low-bit world models and latent imagination losing near-miss outcomes, and how to read driving actions from a frozen world-action model. I also explore world models as decision modules for tool-using language agents. Previously I worked on multimodal reinforcement learning for autonomous driving at the MIP Lab.

Research Interests

Focus
World Models & World-Action Models Model-Based Planning Reinforcement Learning Efficient World Models Reliable World Models Embodied Agents Language Agents

Education

Mar. 2023 —
Feb. 2027
(Expected)

B.S. Candidate in Information and Communication Engineering

Chungbuk National University, School of Information and Communication Engineering · Cheongju, Korea

GPA 3.83 / 4.5

Research Experience

Jun. 2026 —
Present

Research Intern

Artificial Intelligence Laboratory (AI Lab), Chungbuk National University · Advisor: Prof. Keon Myung Lee

Researching efficient and reliable world models and their use in LLM agents.

  • Showed that MPPI planners exploit deterministic rounding error in low-bit TD-MPC2 world models; training-free dithering removed a median 95% of the success drop on Meta-World.
  • Found that latent world-model imagination blurs near misses: success-vs-near-miss AUROC fell from 0.953 to 0.853 on DINO-WM PushT, while far misses stayed separable.
  • Showed that a frozen decision model, JEV, works as an LLM agent's world model without task training, reaching 93.4% success in TextWorld vs. 79.5% for a failure-memory baseline.
Mar. 2025 —
Jun. 2026

Research Intern

Multimedia Information Processing (MIP) Laboratory, Chungbuk National University · Advisor: Prof. Hyun Soo Kang

Conducted research on multimodal reinforcement learning for autonomous driving.

  • Designed and implemented RCS-AC, a CARLA driving actor-critic that gates RGB, LiDAR, route, and ego-state features by sensor reliability, risk, and Shapley contribution.

Activities & Leadership

Oct. 2026 —
Present

Runner (13th Cohort), Hello, World! Season 2 — World Models Paper Review

Pseudo Lab (가짜연구소)

Reviewing world-model research papers, participating in technical discussions, and contributing structured review documents to the project website.

Jan. 2026 —
Sep. 2026

Team Leader, Applied Machine Learning Division

HyperCore AI & Data Analytics Club

Formed and led a club team for hackathons and paper reviews.

Publications & Manuscripts

Under Review

  1. KAIA 2026

    Planners Exploit Deterministic Error: Dithered Low-Bit World Models

    Song, YongHwi* and Lee, Keon Myung

    Submitted to the Korean Artificial Intelligence Association (KAIA) Conference (poster), 2026. Extended version to be submitted to Robotics: Science and Systems (RSS) 2027. · Code

  2. KAIA 2026

    Near-Miss Blindness: Latent World-Model Imagination Loses Resolution at the Outcome Boundary

    Song, YongHwi* and Lee, Keon Myung

    Submitted to the Korean Artificial Intelligence Association (KAIA) Conference (poster), 2026. Extended version to be submitted to Transactions on Machine Learning Research (TMLR). · Code

  3. KAIA 2026

    Imagine Once, Act Fast: Lightweight Action Readout from a Frozen Driving World-Action Model

    Song, YongHwi* and Kang, Hyun Soo

    Submitted to the Korean Artificial Intelligence Association (KAIA) Conference (poster), 2026. · Code

In Preparation

  1. TMLR

    How Many Candidates Does a One-Step World Model Need? A Random-Codebook Account of Nearest-of-m Training on Frozen DINO Tokens

    Song, YongHwi*

    To be submitted to Transactions on Machine Learning Research (TMLR). · Code

  2. EACL SRW 2027

    K-PragShift: Diagnosing Korean Pragmatic Revisions in Tool-Using Language Models

    Song, YongHwi*

    To be submitted to the EACL 2027 Student Research Workshop.

  3. TMLR

    Imagine Once, Act Fast: Lightweight Action Readout from a Frozen Driving World-Action Model

    Song, YongHwi* and Lee, Keon Myung

    To be submitted to Transactions on Machine Learning Research (TMLR). · Code

Conference Presentations

  1. KIIS 2026

    A World Model for LLM Agents: Comparing a Decision Model with General-Purpose LLMs

    Song, YongHwi* and Lee, Keon Myung

    Oral presentation, 2026 Fall Conference of the Korean Institute of Intelligent Systems (KIIS). · Code

Additional Publications

  1. 2020

    A Comparative Study on the Prevalence of Forensic Flies Using Chicken Corpse

    Hyeban Namgung, Kuk Seorin, Park Geonwoo, Bae Seohyun, Song Yonghwi, Lee Hyesong, Han Jian, Hyojoong Kim

    Journal of Science Education for the Gifted, 2020.

* First author.

Projects

Sep. 2026 —
Present

Moonkeeper — CLPS Lunar Mission Explorer

2026 NASA Space Apps Challenge Team Project

Building a React web app that visualizes CLPS lunar landing sites, Sun/Earth visibility windows, and mission timelines from NASA data.

Jul. 2026

Folding — Local-First Document Agent

Team Project

Built the initial LLM-wiki platform that turns HWP, Office, and PDF files into a linked knowledge base.

May 2026

Solar TutorBoard: AI-Powered Personalized Learning Platform

3rd MixUp AI Hackathon (Upstage × BDAI)

Built AI-agent workflows and backend infrastructure for an educational AI platform.

Mar. 2026 —
Jun. 2026

Vehicle and Pedestrian Trajectory Forecasting on Argoverse 2

Personal Project

Compared linear, LSTM, Transformer, and diffusion trajectory forecasters; PCA latent diffusion reached a minFDE of 0.86 at K=16.

Sep. 2025 —
Present

Comparative Survey of RL Research for Autonomous Driving

Independent Research Project

Analyzing 93 RL-based autonomous driving studies and summarizing key research trends.

Jun. 2025 —
Jun. 2026

Shapley-guided Multimodal RL for Autonomous Driving

Faculty-Supervised Research Project

Implemented Shapley-guided multimodal fusion using RGB, LiDAR, route, and ego-state inputs.

Competitions & Hackathons

Nov. 2026
(Upcoming)

2026 NASA Space Apps Challenge — Seoul Local Event

Accepted as a participant in the Seoul Local Event with Team Moonkeeper.

Oct. — Nov.
2026

2026 Southeast Region Multi-Domain Mobility Business Hackathon

Finalist · SeaSight · Selected as one of the top 10 teams for the final round.

Built a drift world model and drone search planner to find fishers who fall overboard; on 88 real drifter cases, the model cut the search area for 90% containment 2.7×.

Oct. 2026
(In Progress)

SK hynix AI Hackathon 2026

Passed the document screening round.

Oct. 2026

KRAFTON AI R&D Hackathon — Dream It Yourself, Video-Based Lightweight World-Action Model

Advanced to the final round.

Built a world-action model that predicts cart–double-pendulum dynamics and selects the next action from the same representation.

Oct. 2026

2026 Startup for All Project (모두의 창업) — Ministry of SMEs and Startups

Folding Next · Passed the first round.

A desktop agent that helps public servants fill forms and draft documents from in-house files, with user-checked sources.

Aug. — Sep.
2026

LG Aimers 9th Cohort — KBO Pitch Control Success Probability Prediction AI Hackathon

Advanced to the Phase 2 online hackathon (top 20%).

Worked on predicting pitch-control success probabilities from game context, player history, and tracking data.

Jul. — Sep.
2026

2026 National Institute of Korean Language AI-Malpyung Competition — Argumentative Writing Scoring

Advanced to the final round.

Developed an AI system for automated evaluation of Korean argumentative writing.

Jul. 2026

2026 CODEGATE AI Startup Hackathon

Finalist · Selected with Folding for the offline final and demo day (about 20 teams).

May 2026

3rd MixUp AI Hackathon (Upstage × BDAI)

Finalist · Solar TutorBoard

Built the agent workflows and backend of a Solar Pro3-based tutoring assistant for lesson reports, payment reminders, and scheduling.

Awards & Scholarships

Mar. 2026

Academic Excellence Scholarship

Chungbuk National University

Sep. 2025 —
Mar. 2026

Undergraduate Research Assistant Scholarship

Chungbuk National University

Mar. 2024

Academic Excellence Scholarship

Chungbuk National University

Sep. 2023

Academic Excellence Scholarship

Chungbuk National University

Skills & Tools

Programming
PythonC/C++Java
ML Frameworks
PyTorchJAXTritonHugging Face Transformersscikit-learnStable-Baselines3
World Models & Simulation
TD-MPC2DreamerV3DINO-WMMuJoCoMeta-WorldLIBEROCARLA
Tools
LinuxGitHydravLLMFastAPI