Students from the Department of Industrial Engineering at UNIST earned two awards at the  2026 K-DATA SCIENCE Conference for research advancing decision-focused learning and synthetic data generation. 

Junhyeong Lee and Sangjin Jin received the Deputy Prime Minister and Minister of Science and ICT Award for “Recovering Predictions without Regretting Decisions: A Cone-constrained Prediction Recovery for Decision-Focused Learning.” Inwoo Tae and Kangmin Kim received the Creative Research Award for “Fix the Timeline First: A Plug-and-Play Retrofit for Sequential Tabular Generators.”

Hosted by the Ministry of Science and ICT (MSIT) and organized by the National Research Foundation of Korea (NRF), the conference was held in Daegu on September 11, 2026. 

Junhyeong Lee and Sangjin Jin focused on a trade-off in decision-focused learning (DFL): improving downstream decision performance can reduce predictive accuracy. Their method recovers prediction accuracy achieved, while preserving the decision performance already.

Inwoo Tae and Kangmin Kim addressed a different challenge in generating sequential synthetic tabular data where existing models can struggle to preserve temporal patterns and relationships within the original data. They developed a plug-and-play correction method that can be applied across different generative models without redesigning the underlying model.

“It is especially rewarding to see student-led research recognized for its academic value,” said Professor Yongjae Lee of the Department of Industrial Engineering, who supervised the research. “We will continue to support students as they take on new research challenges.”