Performance comparison of intermediate-task trained models on XTREME-R and XTREME benchmarks across model sizes

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: How does the performance of intermediate-task trained models on the XTREME-R benchmark compare to their performance on the original XTREME benchmark when using models of varying sizes (e.g., 7B vs. 13B vs. 70B parameters) in zero-shot cross-lingual transfer settings?

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