Data Intelligence System Lab (DISLab)
Department of Industrial and Systems Engineering & Graduate School of Data Science, KAIST, South Korea
Welcome to DISLab! We are dedicated to building trustworthy Agentic AI systems grounded in data-centric intelligence. We believe that intelligence does not emerge from model scale alone, but from how data is perceived, structured, remembered, and used throughout an agent's lifecycle. We studies how data quality, data structure, and data utilization fundamentally shape an agent’s ability to reason, act, and adapt in complex real-world environments. Our goal is to design AI agents that can operate reliably under noisy inputs, long contexts, and imperfect information.
To this end, our research is organized around four core components of Agentic AI:
-
Agent Perception: Understanding multi-modal and long-context signals across text, images, video, and other sensory inputs to perceive complex environments. -
Agent Intelligence: Developing reasoning and decision-making capabilities that integrate world understanding with value alignment to enable purposeful and safe agent behavior. -
Agent Memory: Designing memory and information compression mechanisms that organize, store, and retrieve knowledge over long horizons for consistent agent behaviors. -
Agent Modeling: Building efficient and controllable agent models through knowledge distillation, compression, safety guardrails, and automatic evaluatio.
인턴십은 상시 모집이며, 관심 있는 학생은 아래 양식을 작성하고 저에게 이메일 보내주세요 (Internships are offered on a rolling basis. Interested students are encouraged to complete the form below and email it to me). [Link].
We are not recruiting MS/PhD students for Fall 2026, as all positions have been filled. If you are interested in joining our lab for graduate studies, we strongly encourage you to apply for an internship first, as priority will be given to those with prior research experience in our group.
You can catch up our recent research interest on our Youtube channel.
News
| Oct 25, 2025 |
A paper on Robust Retrieval-augmented Generation was accepted at WSDM 2026.
|
|---|---|
| Aug 21, 2025 |
Two papers on Long Document Comprehension and Recommender System using LLMs were accepted at EMNLP 2025.
|
| Aug 6, 2025 |
Hwanjun was invited as a Keynote Speaker at the International Workshop on Retrieval-Driven Generative AI & ScienceON AI Challenge, held as part of the CIKM conference.
|
| Jul 21, 2025 |
Hwanjun was nominated as an Area Chair at AAAI 2026, responsible for handling 100 submissions.
|
| Jul 8, 2025 |
A paper on Reflective Reasoning for Text Summarization was accepted at COLM 2025.
|
| Jun 26, 2025 |
A paper on Data Condensation with Noisy Labels was accepted at ICCV 2025.
|
| Jun 21, 2025 |
Our two MS students, Yuho Lee and Taewon Yun, successfully passed their MS dissertation defense. Congratulations to both of them!
|