2 papers
q-bio.GN2026
DNACHUNKER: Learnable Tokenization for DNA Language Models
Taewon Kim, Jihwan Shin, Hyomin Kim +5
DNA language models are increasingly used to represent genomic sequence, yet their effectiveness depends critically on how raw nucleotides are converted into model inputs. Unlike n…
cs.LG2025
Heterogeneous Federated Learning with Prototype Alignment and Upscaling
Gyuejeong Lee, Jihwan Shin, Daeyoung Choi
Heterogeneity in data distributions and model architectures remains a significant challenge in federated learning (FL). Various heterogeneous FL (HtFL) approaches have recently bee…