Scaling of English Intermediate-Task Fine-Tuned Models on Non-English XTREME-R Reasoning Tasks

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 performance of English intermediate-task fine-tuned models scale with model size (100M to 10B+ parameters) on non-English reasoning tasks in XTREME-R, measured by accuracy improvements?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.9/10.