Per-token inference latency comparison between intermediate-task training and zero-shot cross-lingual transfer across model sizes

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 per-token inference latency compare between English intermediate-task training and direct zero-shot cross-lingual transfer when using the same model architecture but varying model sizes (e.g., 7B vs 13B vs 30B parameters) on the XGLUE benchmark?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.3/10.