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 tasResearch goal: Does the choice of intermediate-task dataset size (e.g., small vs. large-scale) affect the efficiency (inference throughput) of zero-shot cross-lingual transfer in LXMERT on tasks like XTREME-R, and by how much does accuracy degrade with smaller intermediate datasets?Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.0/10.