Intermediate-Task Training Effects on Multilingual Model Performance Across Domains

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 intermediate-task training with English data affect the performance of multilingual models like mBERT or XLM-R on the XTREME benchmark when the intermediate task is from a different domain (e.g., natural language inference vs. question answering) compared to the target task?

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