Analyzing Bagging Methods for Language Models

Pranab Islam, Shaan Khosla, Arthur Lok +1

Modern language models leverage increasingly large numbers of parameters to achieve performance on natural language understanding tasks. Ensembling these models in specific configu…