XTREME-R F1 Score Variations with Intermediate Task Counts in English 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: What is the impact of varying the number of intermediate tasks (3 vs. 9) on XTREME-R F1 scores when using English intermediate-task training, measured in terms of computational efficiency (inference latency) and accuracy trade-offs?

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