Kyungpook National University Β· School of CSE
We build generative language models that reason over code, use tools, and work as reliable coding agents β and we study what makes LLMs behave the way they do.
Prof. Seonghyeon Lee has been invited to serve as a senior program committee member for AAAI-27.
A new government-funded project, Development of Automation and Knowledge Accumulation Technology for Industrial Data-Driven Physics-Informed Neural Networks, has started (PI: Prof. Suyeon Kim).
Our paper Layer-wise Target Propagation: Efficient Component Attribution through Target Centric Propagation has been accepted to the EMNLP 2026 main conference.
Prof. Suyeon Kim has joined our lab. Welcome!
Prof. Seonghyeon Lee has been invited as a reviewer for ACM Transactions on Software Engineering and Methodology (TOSEM, Q1).
Methods that let language models reason over code, interact with external tools, and use structured context to solve complex programming tasks β improving the reliability, controllability, and practical usefulness of code-generating agents in real development settings.
LLM-based methods and analyses that explain how language models behave, and that strengthen their capabilities through both training and inference-time techniques.
μΈκ³΅μ§λ₯ 보쑰 μννΈμ¨μ΄ κ°λ°μ μν μ½λ μ λ¬Έκ° μΈμ΄λͺ¨λΈ λ° μ½λ© μμ΄μ νΈ κ°λ°
Code-expert language models and coding agents that understand and operate over entire code repositories: repository-level code completion, context construction across multiple files, long-range dependency modeling in code, and agentic workflows in which models navigate large codebases, reason about project structure, and iteratively interact with tools.
μ°μ λ°μ΄ν° κΈ°λ° λ¬Όλ¦¬μ 보 μ κ²½λ§ μλν λ° μ§μ μ¬μ¬μ© κΈ°μ κ°λ°
Layer-wise Target Propagation: Efficient Component Attribution through Target Centric Propagation
Acceptance rate: 15.4%
On the Effectiveness of Integration Methods for Multimodal Dialogue Response Retrieval
The 16th ACM International Conference on Multimedia Retrieval
Dynamic Multi-period Experts for Online Time Series Forecasting
Proceedings of The ACM Web Conference 2026
Harmonic Dataset Distillation for Time Series Forecasting
Proceedings of the 40th AAAI Conference on Artificial Intelligence
