mT5 Performance in Code-Mixed Text via Intermediate Code Task Training

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 intermediate-task training with code-focused tasks (e.g., Python, Java) affect zero-shot cross-lingual transfer performance for code-mixed text (e.g., Spanish-English) in mT5, evaluated on XTREME-R with accuracy and F1 metrics? Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.2/10.