Domain Shift Between English Intermediate and Non-English Target Tasks in XTREME-R Zero-Shot 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: What is the impact of domain shift between English intermediate tasks and non-English target tasks on cross-lingual zero-shot performance in the XTREME-R benchmark, and how can domain alignment mitigate this effect?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.9/10.