BabyHGRN: Exploring RNNs for Sample-Efficient Training of Language Models

TL;DR

This paper demonstrates that RNN-based models like BABYHGRN can outperform transformer models in low-resource language modeling tasks, especially when combined with knowledge distillation.

Contribution

It introduces BABYHGRN, an RNN-based language model that outperforms transformers in low-resource settings, challenging the dominance of transformer architectures.

Findings

BABYHGRN outperforms transformers on multiple benchmarks.

Knowledge distillation improves RNN model performance.

RNNs are viable alternatives in resource-constrained environments.

Abstract

This paper explores the potential of recurrent neural networks (RNNs) and other subquadratic architectures as competitive alternatives to transformer-based models in low-resource language modeling scenarios. We utilize HGRN2 (Qin et al., 2024), a recently proposed RNN-based architecture, and comparatively evaluate its effectiveness against transformer-based baselines and other subquadratic architectures (LSTM, xLSTM, Mamba). Our experimental results show that BABYHGRN, our HGRN2 language model, outperforms transformer-based models in both the 10M and 100M word tracks of the challenge, as measured by their performance on the BLiMP, EWoK, GLUE and BEAR benchmarks. Further, we show the positive impact of knowledge distillation. Our findings challenge the prevailing focus on transformer architectures and indicate the viability of RNN-based models, particularly in resource-constrained environments.