Zero-Shot Cross-Lingual Transfer in XTREME via Intermediate Training on High-Resource Non-English Languages

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 does the performance of zero-shot cross-lingual transfer in XTREME tasks vary when using intermediate-task training on high-resource non-English languages (e.g., Spanish, Chinese, German) compared to English, measured by accuracy and adversarial robustness?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.7/10.