Multimodal Intermediate Tasks for Zero-Shot Cross-Lingual Transfer on XTREME-R

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: Do multimodal intermediate tasks (e.g., VLM-based tasks) improve zero-shot cross-lingual transfer performance on XTREME-R compared to text-only intermediate tasks?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.1/10.