Intermediate-Task Fine-Tuning for Cross-Lingual Transfer in Multimodal Models

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: Does English intermediate-task fine-tuning improve zero-shot cross-lingual transfer in multimodal models (e.g., CLIP or BLIP) when evaluated on benchmarks like MMMU or VQAv2, as measured by accuracy or F1 score?

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