Zero-Shot Cross-Lingual Transfer in Multimodal vs. Text-Only Models: Intermediate-Task Training Effects

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 intermediate-task training on zero-shot cross-lingual transfer when using multimodal models (e.g., Flamingo or PaLI) compared to text-only models on benchmarks like XTREME-R, and how does this compare to the performance of models like mT5?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.0/10.