Paper · arXiv 2410.24159
GPT or BERT: why not both?
Lucas Georges Gabriel Charpentier, David Samuel
A hybrid model combining masked and causal language modeling within a transformer stack outperforms models using either paradigm alone.
Abstract
We present a simple way to merge masked language modeling with causal language modeling. This hybrid training objective results in a model that combines the strengths of both modeling paradigms within a single transformer stack: GPT-BERT can be transparently used like any standard causal or masked language model. We test the pretraining process that enables this flexible behavior on the BabyLM Challenge 2024. The results show that the hybrid pretraining outperforms masked-only or causal-only models. We openly release the models, training corpora and code.