Performance Gains of Intermediate-Task Training on XTREME-R Across Model Scales

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 the performance gain from English intermediate-task training on XTREME-R persist when scaling from 3B to 7B parameter models, and how does this compare to direct target-task fine-tuning in terms of F1 scores?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.7/10.