Scaling Intermediate-Task Data Size and Robustness in Zero-Shot Cross-Lingual Transfer of 70B-Parameter Models

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 scaling of intermediate-task training data size (e.g., 1K vs. 100K samples per task) influence the robustness of a 70B-parameter model's zero-shot cross-lingual transfer on XTREME, measured by accuracy stability across low-, medium-, and high-resource languages?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 7.7/10.