Comparative Analysis of English versus Multilingual Intermediate Training for Zero-Shot Cross-Lingual XTREME-R Performance

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 zero-shot cross-lingual performance of models trained on English intermediate tasks compare to multilingual intermediate-task training on downstream XTREME-R benchmarks, measured by task-specific accuracy?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.6/10.