Pretraining Language Models with LoRA and Artificial Languages

AbstractLarge language models (LLMs) require a substantial amount of training data, which contrasts with the data-efficient learning observed in humans. In our submission to the BabyLM Challenge, we address this disparity by proposing a parameter-efficient pretraining approach for language acquisition from limited data. Our approach involves initializing the model with token embeddings trained by a shallow model, followed by tuning the non-embedding parameters with non-linguistic data to introduce structural biases. Then, we freeze the resulting model and pretrain it on the 10M-token BabyLM corpus using LoRA adapters. Experiments on small corpora demonstrate that our approach improves upon classic pretraining of the entire model.