Cross-lingual Few-shot Performance with Mixed Monolingual and Multilingual Intermediate-task Training

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: Does incorporating a mix of monolingual and multilingual intermediate-task training (e.g., using the XTREME-R benchmark) improve few-shot cross-lingual performance on structured prediction tasks compared to English-only intermediate training, as measured by F1 scores on the target languages?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.0/10.