Intermediate Task Difficulty and 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) in multilingual settings affect zero-shot cross-lingual transfer performance on XTREME-R, and can task ranking based on difficulty improve model generalization?

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