Impact of Intermediate Task Difficulty on Zero-Shot Cross-Lingual Transfer Performance in 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: How does the choice of intermediate task difficulty (measured by task-specific accuracy) impact zero-shot cross-lingual transfer performance on the XTREME-R benchmark when using English-only intermediate tasks compared to multilingual ones, as evaluated by Spearman correlation between task difficulty and transfer accuracy?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.4/10.