Multilingual Intermediate-Task Transfer in XTREME Benchmark 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: Does intermediate-task transfer from English tasks to non-English target tasks in XTREME still hold when the intermediate task is a multilingual task (e.g., XNLI) rather than a monolingual English task, and how does this affect performance on XTREME's UDA and TYDI QA benchmarks?

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