Inference Efficiency Trade-off in Zero-shot Cross-lingual Transfer with mT5-large

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 inference efficiency trade-off (measured in tokens/sec) when using mT5-large with intermediate-task training for zero-shot cross-lingual transfer on MMLU compared to direct fine-tuning? Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.2/10.