Scaling Effects of Model Size on Zero-Shot Cross-Lingual Accuracy and Computational Efficiency in Monolingual vs. Multilingual

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 scaling of model size (1B to 20B parameters) affect zero-shot cross-lingual accuracy on XTREME-R and XTREME-P when using monolingual English intermediate-task training compared to multilingual training, and what is the computational efficiency trade-off in terms of training time per parameter?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.2/10.