Adversarial Robustness in Zero-Shot Cross-Lingual Transfer Models Post-English QA Training

Abstract

Intermediate-task training---fine-tuning a pretrained model on an intermediate task before fine-tuning again on the target task---often improves model performance substantially on language understanding tasks in monolingual English settings. We investigate whether English intermediate-task training is still helpful on non-English target tasks. Using nine intermediate language-understanding tasks, we evaluate intermediate-task transfer in a zero-shot cross-lingual setting on the XTREME benchmark. We see large improvements from intermediate training on the BUCC and Tatoeba sentence retrieval tas Research goal: How robust are zero-shot cross-lingual transfer models (e.g., XLM-R, mT5) to adversarial examples in target languages after intermediate training on English question-answering tasks, measured by accuracy degradation on perturbed XTREME-R test sets? Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.3/10.