Harnessing Dataset Cartography for Improved Compositional Generalization in Transformers

Osman Batur İnce, Tanin Zeraati, Semih Yagcioglu, Yadollah Yaghoobzadeh, Erkut Erdem, Aykut Erdem

Submission Type: Regular Long Paper

Submission Track: Machine Learning for NLP

Submission Track 2: Efficient Methods for NLP

Keywords: dataset cartography, compositional generalization, training dynamics

TL;DR: Leveraging dataset cartography improves compositional generalization in Transformer models, boosting accuracy by up to 10% on CFQ and COGS datasets without hyperparameter tuning.

Abstract: Neural networks have revolutionized language modeling and excelled in various downstream tasks. However, the extent to which these models achieve compositional generalization comparable to human cognitive abilities remains a topic of debate. While existing approaches in the field have mainly focused on novel architectures and alternative learning paradigms, we introduce a pioneering method harnessing the power of dataset cartography (Swayamdipta et al., 2020). By strategically identifying a subset of compositional generalization data using this approach, we achieve a remarkable improvement in model accuracy, yielding enhancements of up to 10% on CFQ and COGS datasets. Notably, our technique incorporates dataset cartography as a curriculum learning criterion, eliminating the need for hyperparameter tuning while consistently achieving superior performance. Our findings highlight the untapped potential of dataset cartography in unleashing the full capabilities of compositional generalization within Transformer models.

Submission Number: 3109