Data Augmentations for Data-Constrained Language Model Pretraining
For AI labs facing data scarcity relative to compute, this work provides a practical regularization method to enable effective multi-epoch training on fixed corpora.
This paper addresses overfitting in autoregressive language model pretraining under data-constrained, compute-abundant conditions. It shows that data augmentations (token noise, sequence permutations, target offsets) reduce validation loss, with random token replacement achieving the best minimum loss among individual methods, and combining categories further improves results.
As AI labs approach a data ceiling where compute capacity outpaces the rate of new high-quality text generation, language model pretraining is shifting toward a data-constrained, compute-abundant regime that demands productive multi-epoch training on fixed corpora. Standard autoregressive (AR) pretraining overfits severely in this setting, reaching its optimum early and then continuously deteriorating. We investigate data augmentation as a regularizer to mitigate this overfitting and enable productive training for hundreds of epochs on the same data. We introduce three orthogonal categories of augmentation for AR pretraining: token-level noise (masking, random replacement), sequence permutations (right-to-left prediction, Fill-in-the-Middle), and target offset prediction ($x_{t+i}$ for $i > 1$). Through systematic ablations, we find that individual augmentations delay overfitting and lower validation loss relative to the baseline, with random token replacement achieving the best minimum loss among individual methods. Combining augmentation categories further lowers the minimum validation loss. Our experiments demonstrate that data augmentations mitigate AR pretraining's data inefficiency and offer a promising solution to the data-constrained regime. All code and data are available at https://github.com/michaelchen-lab/data-augmentations-for-pretraining