3 papers
cs.LG2026
Bastion: Budget-Aware Speculative Decoding with Tree-structured Block Diffusion Drafting
Soowon Oh, Nam Cao, Yujin Kim +4
Block-diffusion drafters have recently emerged as a powerful alternative for speculative decoding by predicting multiple future-token distributions in a single parallel step. Howev…
cs.CL2026
PerMix-RLVR: Preserving Persona Expressivity under Verifiable-Reward Alignment
Jihwan Oh, Soowon Oh, Murad Aghazada +3
Persona prompting has been widely adopted to steer large language models (LLMs) behavior and improve their instruction performance by assigning specific characters. However, identi…
cs.LG2026
mSFT: Addressing Dataset Mixtures Overfitting Heterogeneously in Multi-task SFT
Woosung Koh, Jeyoung Jeon, Youngjin Song +4
Current language model training commonly applies multi-task Supervised Fine-Tuning (SFT) using a homogeneous compute budget across all sub-datasets. This approach is fundamentally…