Sihyun Ha (Integrated M.S.–Ph.D. student) from the Department of Electrical Engineering at POSTECH (advisor: Prof. Yosep Jeon) received the Outstanding Paper Award at the 2025 Summer Conference of the Korean Institute of Communications and Information Sciences (KICS) for his paper titled “A LoRA Method for Fine-Tuning in Heterogeneous Federated Learning Environments.”

The KICS Summer Conference is one of the largest academic conferences in Korea in the field of electrical and electronic engineering, with more than 2,000 participants attending each year.

This research addresses the rank heterogeneity problem that arises when different clients in a federated learning environment employ LoRA (Low-Rank Adaptation) modules with varying ranks. To resolve this issue, the study proposes a Rank-Aware Aggregation method that applies rank-level weighting during the model aggregation process.

The proposed method decomposes LoRA modules in a rank-wise manner and designs aggregation weights that reflect both the dataset size and rank information of each client, enabling more efficient and effective model integration across heterogeneous clients.