Intermediate-Task Training with Synthetic Data for Robust Zero-Shot Cross-Lingual Text Classification

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: Does intermediate-task training with synthetic data (e.g., back-translated or paraphrased samples) improve zero-shot cross-lingual transfer robustness for text classification in low-resource languages compared to native intermediate-task training, as measured by accuracy degradation under adversarial noise in the XTREME-R benchmark?

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