Domain-Specific Intermediate-Task Training Effects on Zero-Shot Cross-Lingual Evaluation Accuracy in XTREME Across Model Scales

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 domain-specific intermediate-task training in English affect the accuracy of zero-shot cross-lingual evaluation on XTREME when scaling up the model size, measured by comparing performance on small vs. large LLMs?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.7/10.