Model Balancing Helps Low-data Training and Fine-tuning

Recent advances in foundation models have emphasized the need to align pre-trained models with specialized domains using small, curated datasets. Studies on these foundation models underscore the importance of low-data training and fine-tuning. This topic, well-known in natural language processing (NLP), has also gained increasing attention in the emerging field of scientific machine learning (SciML). To address the limitations of low-data training and fine-tuning, we draw inspiration from Heavy-Tailed Self-Regularization (HT-SR) theory, analyzing the shape of empirical spectral densities (ESDs) and revealing an imbalance in training quality across different model layers. To mitigate this issue, we adapt a recently proposed layer-wise learning rate scheduler, TempBalance, which effectively balances training quality across layers and enhances low-data training and fine-tuning for both NLP and SciML tasks. Notably, TempBalance demonstrates increasing performance gains as the amount of available tuning data decreases. Comparative analyses further highlight the effectiveness of TempBalance and its adaptability as an "add-on" method for improving model performance. * Equal contribution. Work completed during an internship at Dartmouth College. Imbalanced Before Using TempBalance Smaller PL_Alpha_Hill Schedule Smaller LR Larger PL_Alpha_Hill Schedule Larger LR PL_Alpha_Hill Layer/Block Index ESD of weight matrices PL_Alpha_Hill Layer/Block Index RoBERTa-base accuracy on 0.2% SST-2: 58.49 RoBERTa-base accuracy on 0.2% SST-2: 68.39 Less Heavy-tailed More Heavy-tailed More Balanced After Using TempBalance