Generalization Measures for Zero-Shot Cross-Lingual Transfer

TL;DR

This paper investigates various measures, including a novel sharpness algorithm, to evaluate and improve the understanding of a language model's ability to generalize in zero-shot cross-lingual transfer scenarios.

Contribution

It introduces a new stable method for measuring sharpness in loss landscapes, enhancing the assessment of model generalization in multilingual settings.

Findings

Sharpness correlates with cross-lingual transfer success.

Variance and distance metrics provide insights into model generalization.

Proposed sharpness algorithm is reliable and computationally efficient.

Abstract

A model's capacity to generalize its knowledge to interpret unseen inputs with different characteristics is crucial to build robust and reliable machine learning systems. Language model evaluation tasks lack information metrics about model generalization and their applicability in a new setting is measured using task and language-specific downstream performance, which is often lacking in many languages and tasks. In this paper, we explore a set of efficient and reliable measures that could aid in computing more information related to the generalization capability of language models in cross-lingual zero-shot settings. In addition to traditional measures such as variance in parameters after training and distance from initialization, we also measure the effectiveness of sharpness in loss landscape in capturing the success in cross-lingual transfer and propose a novel and stable algorithm to reliably compute the sharpness of a model optimum that correlates to generalization.