English

Critical Learning Periods: Leveraging Early Training Dynamics for Efficient Data Pruning

Computation and Language 2024-06-24 v2

Abstract

Neural Machine Translation models are extremely data and compute-hungry. However, not all data points contribute equally to model training and generalization. Data pruning to remove the low-value data points has the benefit of drastically reducing the compute budget without significant drop in model performance. In this paper, we propose a new data pruning technique: Checkpoints Across Time (CAT), that leverages early model training dynamics to identify the most relevant data points for model performance. We benchmark CAT against several data pruning techniques including COMET-QE, LASER and LaBSE. We find that CAT outperforms the benchmarks on Indo-European languages on multiple test sets. When applied to English-German, English-French and English-Swahili translation tasks, CAT achieves comparable performance to using the full dataset, while pruning up to 50% of training data. We inspect the data points that CAT selects and find that it tends to favour longer sentences and sentences with unique or rare words.

Keywords

Cite

@article{arxiv.2405.19462,
  title  = {Critical Learning Periods: Leveraging Early Training Dynamics for Efficient Data Pruning},
  author = {Everlyn Asiko Chimoto and Jay Gala and Orevaoghene Ahia and Julia Kreutzer and Bruce A. Bassett and Sara Hooker},
  journal= {arXiv preprint arXiv:2405.19462},
  year   = {2024}
}

Comments

Accepted to ACL 2024 Findings

R2 v1 2026-06-28T16:46:18.137Z