Zero-shot accuracy comparison on XTREME-M with English intermediate-task training versus direct fine-tuning for low-resource

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 accuracy on XTREME-M compare when using English intermediate-task training data at 50M scale versus direct target-task fine-tuning when evaluated on low-resource languages with less than 1M training examples?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.1/10.