Multimodal Model Performance: Intermediate Task Training with Cross-Lingual Data and Zero-Shot Evaluation 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: For multimodal models, how does combining intermediate task training with English text and non-English image captions affect zero-shot performance on XTREME-R versus monolingual English text-only training, evaluated by F1 score on cross-lingual NLI tasks?

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