Data Intelligence System Lab. (DISL)
Department of Industrial and Systems Engineering & Graduate School of Data Science, KAIST, South Korea
E2 #3110, 291 Daehak-ro, Yuseong-gu, Daejeon, South Korea
Welcome to the DISL Lab. We are dedicated to pioneering advancements in the field of artificial intelligence (AI). Our vision is centered on the pursuit of data-centric approaches that enhance the perofrmance of AI algorithms and systems.
The foundation role of data remains unwavering, even as AI trends evolve rapidly. Our research interests involve making innovations with cutting-edge technologies in various domains, including computer vision (CV) and natural language processing (NLP). Our research scope is expansive and adapts with the advancements in AI. Currently, our primary focus lies in tackling novel challenges in data for natural language generation, including how to evaluate LLM’s outputs, how to expedite the Transformer’s inference, and etc. Additionally, we are actively exploring data-robust and data-efficient AI modeling, including learning with imperfect data, and AI traning and inference under real-world setup, which encompasses continual learning and online adaptation.
Please see the Join tab if you are interested in applying for Intern, MS, PhD, and PostDoc positions.
You can catch up our recent research interest on our Youtube channel.
News
| Apr 2, 2024 |
Hwanjun was honored to be selected as a fellow in the Asian Trustworthy Machine Learning (ATML) group and received the prestigious ATML Fellowship from the two hosts, Prof. Bo Han and Prof. Tongliang Liu.
|
|---|---|
| Mar 15, 2024 |
Three research papers on multimodal data creation, hallucination evaluation, semi-supervised text summarization was accepted at the main track of NAACL (long paper) as two Oral and one Poster Papers.
|
| Mar 9, 2024 | A research paper (collaboration with AWS AI Labs) on unravelling challenges with intent encoders was released, Can Your Model Tell a Negation from an Implicature? Unravelling Challenges With Intent Encoders. |
| Feb 21, 2024 | A research paper (collaboration with AWS AI Labs) on creating multimodal dialouge datasets was released, MAGID: An Automated Pipeline for Generating Synthetic Multi-modal Datasets. |
| Feb 21, 2024 | A research paper (collaboration with AWS AI Labs) on hallucination evaluation was released, ToFuEval: Evaluating Hallucinations of LLMs on Topic-Focused Dialogue Summarization. |
| Jan 16, 2024 |
We received a silver price at the Samsung Humantech Paper Award (Signal Processing / NLP / Paper: Robust Early-exiting)
|
| Jan 11, 2024 |
A research paper on Time-series Anomaly Detection accepted at TheWebConf (WWW) 2024
|