4 papers
Benchmarks Are Not That Out of Distribution: Word Overlap Predicts Performance
Woojin Chung, Jeonghoon Kim
Understanding what constitutes high-quality pre-training data remains a central question in language model training. In this work, we investigate whether benchmark performance is p…
Exploiting Vocabulary Frequency Imbalance in Language Model Pre-training
Woojin Chung, Jeonghoon Kim
Large language models are trained with tokenizers, and the resulting token distribution is highly imbalanced: a few words dominate the stream while most occur rarely. Recent practi…
On the Robustness of Reward Models for Language Model Alignment
Jiwoo Hong, Noah Lee, Eunki Kim +5
The Bradley-Terry (BT) model is widely practiced in reward modeling for reinforcement learning with human feedback (RLHF). Despite its effectiveness, reward models (RMs) trained wi…
Stable Language Model Pre-training by Reducing Embedding Variability
Woojin Chung, Jiwoo Hong, Na Min An +2
Stable pre-training is essential for achieving better-performing language models. However, tracking pre-training stability by calculating gradient variance at every step is impract…