Task-specific vs. general intermediate training in zero-shot cross-lingual performance for 100M--1B models on XTREME-R

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 impact of using task-specific intermediate training (e.g., NLI, QA) versus general intermediate training (e.g., filler prediction) on zero-shot cross-lingual performance for 100M vs. 1B parameter models on the XTREME-R benchmark?

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