Enhancing Cross-lingual Transfer by Manifold Mixup

TL;DR

This paper introduces X-Mixup, a novel method that reduces cross-lingual representation discrepancies, leading to improved transfer performance across multiple languages in text understanding tasks.

Contribution

The paper proposes the cross-lingual manifold mixup (X-Mixup) technique, which adaptively calibrates representations to enhance cross-lingual transfer performance.

Findings

X-Mixup achieves 1.8% performance gains on XTREME benchmark tasks.

X-Mixup significantly reduces cross-lingual representation discrepancy.

Improves transfer performance across multiple languages.

Abstract

Based on large-scale pre-trained multilingual representations, recent cross-lingual transfer methods have achieved impressive transfer performances. However, the performance of target languages still lags far behind the source language. In this paper, our analyses indicate such a performance gap is strongly associated with the cross-lingual representation discrepancy. To achieve better cross-lingual transfer performance, we propose the cross-lingual manifold mixup (X-Mixup) method, which adaptively calibrates the representation discrepancy and gives a compromised representation for target languages. Experiments on the XTREME benchmark show X-Mixup achieves 1.8% performance gains on multiple text understanding tasks, compared with strong baselines, and significantly reduces the cross-lingual representation discrepancy.