Zero-shot Transfer Accuracy in XTREME-R: Multimodal vs. Text-Only Intermediate-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 the zero-shot transfer accuracy on XTREME-R compare when using intermediate-task training on multimodal datasets like VQA versus text-only tasks, and what is the impact on alignment between modalities?

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