Multimodal Intermediate-Task Training for Zero-Shot Cross-Lingual Transfer Robustness in Low-Resource Languages

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: What is the impact of incorporating multimodal (text-image) intermediate-task training in English on zero-shot cross-lingual transfer robustness for low-resource languages in XTREME-R, evaluated using downstream task accuracy and cross-lingual consistency scores?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.9/10.