B.S. Candidate · Chungbuk National University
World Models · World Action Models · Agentic AI · Tool-Using Language Models
Academic Homepage · CV (PDF) · Email
I am an undergraduate researcher in Information and Communication Engineering at Chungbuk National University. My research focuses on efficient world and world-action models, agentic AI, and tool-using language models. Previously, I worked on multimodal reinforcement learning for autonomous driving.
| Period | Role & Affiliation |
|---|---|
| Mar. 2023 – Feb. 2027 (Expected) | B.S. Candidate, School of Information and Communication Engineering, Chungbuk National University · GPA 3.83 / 4.5 |
| Jun. 2026 – Present | Research Intern, Artificial Intelligence Laboratory (AI Lab), Chungbuk National University (Advisor: Prof. Keon Myung Lee) · Conducting research on agentic AI, tool-using language models, and efficient world/world-action models. |
| Mar. 2025 – Jun. 2026 | Research Intern, Multimedia Information Processing (MIP) Laboratory, Chungbuk National University (Advisor: Prof. Hyun Soo Kang) · Research on Autonomous Driving and Multimodal Reinforcement Learning. |
| Period | Role & Organization |
|---|---|
| Oct. 2026 – Present | Researcher, Hello, World! Season 2 — World Models Paper Review, Pseudo Lab (가짜연구소) · Reviewing World Models research papers, participating in technical discussions, and contributing structured review documents to the project website. |
| Jan. 2026 – Present | Advanced Division Leader, Applied Machine Learning Division, HyperCore AI & Data Analytics Club · Leading machine learning study sessions and supporting AI project development. |
- World Models and World Action Models
- Agentic AI and Tool-Using Language Models
- Large and Small Language Models (LLMs/sLMs)
- Multimodal Learning and Multimodal Foundation Models
- Shapley-style Contribution-Guided Sensor Fusion for Robust Reinforcement Learning-Based Autonomous Driving
Song, YongHwi* · Submitted to KIISE Transactions on Computing Practices (KTCP), 2026. - A Decision Model as a World Model: Comparing JEV with Generative LLMs for LLM Agents
Song, YongHwi* and Lee, Keon Myung · Submitted to the 2026 Fall Conference of the Korean Institute of Intelligent Systems (KIIS), 2026.
- K-PragShift: Diagnosing Korean Pragmatic Revisions in Tool-Using Language Models
Song, YongHwi* · To be submitted to the EACL 2027 Student Research Workshop. - Lightweight Action Readout from Multi-Level World Representations for Efficient DriveWAM
Song, YongHwi* and Lee, Keon Myung · To be submitted to Transactions on Machine Learning Research (TMLR). - Lightweight Action Readout from Multi-Level World Representations for Efficient DriveWAM
Song, YongHwi* and Lee, Keon Myung · To be submitted to the poster track of the Korean Artificial Intelligence Association (KAIA) Conference. - A Comparative Survey of Domestic and International Reinforcement Learning Research for Autonomous Driving
Song, YongHwi* · Manuscript in preparation, 2026.
* First author. Submitted work is listed with its current submission status.
- Hyeban Namgung, Kuk Seorin, Park Geonwoo, Bae Seohyun, Song Yonghwi, Lee Hyesong, Han Jian, Hyojoong Kim. “A Comparative Study on the Prevalence of Forensic Flies Using Chicken Corpse.” Journal of Science Education for the Gifted, 2020.
Team Project · Jul. 2026
Developed a local-first document agent for retrieving, connecting, and safely editing heterogeneous documents.
Upstage MixUp Agent Hackathon · May 2026
Built AI-agent workflows and backend infrastructure for an educational AI platform.
Personal Project · Mar. 2026 – Jun. 2026
Implemented trajectory prediction models using the Argoverse 2 Motion Forecasting dataset.
Research Project · Jan. 2026 – Present
Investigating efficient action prediction by extracting lightweight action representations from multi-level world representations.
Independent Research Project · Sep. 2025 – Present
Analyzing 93 RL-based autonomous driving studies and summarizing key research trends.
Faculty-Supervised Research Project · Jun. 2025 – Jun. 2026
Implemented Shapley-guided multimodal fusion using RGB, LiDAR, route, and ego-state inputs.
Main implementation · Experiments · Prototype
| Event | Result & Work |
|---|---|
| 2026 NASA Space Apps Challenge (Seoul) · Nov. 2026, upcoming | Accepted as a participant in the Seoul Local Event with Team Moonkeeper. Planning a CLPS lunar mission explorer that visualizes landing sites, Sun/Earth visibility, power and communication windows, and mission timelines from NASA data. |
| KRAFTON AI R&D Hackathon · Oct. 2026, in progress | Participating individually in Problem 3, Dream It Yourself. Developing a lightweight video-based world-action model that predicts future frames and selects actions to swing up and balance a cart–double-pendulum from offline RGB data. |
| LG Aimers 9th Cohort · Aug. – Sep. 2026 | Advanced to Phase 2 Online AI Hackathon (Top 20%). Worked on predicting pitch-control success probabilities from game context, player history, and tracking data. |
| 2026 NIKL AI-Malpyung — Argumentative Writing Scoring · Jul. – Sep. 2026 | Ranked 23rd in the preliminary competition and advanced to the 50-team final arena. Developed an AI system for automated evaluation of Korean argumentative writing. |
| 2026 CODEGATE AI Startup Hackathon · Jul. 2026 | Finalist · Selected as one of approximately 20 finalist teams for the offline hackathon and demo day. Participated in intensive development and validation of an AI-based startup MVP. |
| Upstage × BDAI Agent Development Hackathon · 2026 | Finalist · Solar TutorBoard / Developed a Solar Pro3-based multi-agent platform for tutoring operations, including lesson reports, payment reminders, and schedule coordination. |
- Academic Excellence Scholarship, Chungbuk National University · Mar. 2026, Mar. 2024, Sep. 2023
- Undergraduate Research Assistant Scholarship, Chungbuk National University · Sep. 2025 – Mar. 2026
| Area | Technologies & Methods |
|---|---|
| Programming | Python, C/C++, Java |
| ML Frameworks | PyTorch, TensorFlow, scikit-learn, Stable-Baselines3 |
| World Models & RL | World Models, World Action Models, Reinforcement Learning, Model-Based Reinforcement Learning, Actor-Critic Methods, Multimodal Policy Learning |
| LLMs & Agentic AI | Large/Small Language Models (LLMs/sLMs), Agentic AI, Tool-Using Language Models, Multi-Agent Systems, Structured LLM Outputs, Local LLM Inference |
| Multimodal & Embodied AI | Multimodal Learning, Sensor Fusion, Computer Vision, Embodied AI, Physical AI, Trajectory Prediction |
| Driving & Simulation | CARLA, Argoverse 2, ManiSkill, Motion Forecasting, Autonomous Driving, Robotic Manipulation |
| AI Systems | Git, Linux, FastAPI, Supabase, Ollama, LM Studio, REST APIs |
Homepage · CV · thddydgnl1937@gmail.com
Updated October 6, 2026.