CAMERO: Consistency Regularized Ensemble of Perturbed Language Models with Weight Sharing

TL;DR

CAMERO introduces a weight-sharing ensemble method with perturbations and consistency regularization to improve language model performance while reducing memory costs.

Contribution

It proposes a novel ensemble approach that maintains diversity and regularization through perturbations and shared weights, enhancing generalization.

Findings

Outperforms standard ensembles on GLUE benchmark

Reduces model size significantly while maintaining accuracy

Demonstrates improved generalization with shared-weight perturbations

Abstract

Model ensemble is a popular approach to produce a low-variance and well-generalized model. However, it induces large memory and inference costs, which are often not affordable for real-world deployment. Existing work has resorted to sharing weights among models. However, when increasing the proportion of the shared weights, the resulting models tend to be similar, and the benefits of using model ensemble diminish. To retain ensemble benefits while maintaining a low memory cost, we propose a consistency-regularized ensemble learning approach based on perturbed models, named CAMERO. Specifically, we share the weights of bottom layers across all models and apply different perturbations to the hidden representations for different models, which can effectively promote the model diversity. Meanwhile, we apply a prediction consistency regularizer across the perturbed models to control the variance due to the model diversity. Our experiments using large language models demonstrate that CAMERO significantly improves the generalization performance of the ensemble model. Specifically, CAMERO outperforms the standard ensemble of 8 BERT-base models on the GLUE benchmark by 0.7 with a significantly smaller model size (114.2M vs. 880.6M).