Batch Size Effects on Latency-Alignment Trade-offs in Intermediate-Task Tuning for XTREME-R Zero-Shot Transfer

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 do different batch sizes affect the trade-off between inference latency and alignment score improvements in intermediate-task tuning versus standard fine-tuning for zero-shot transfer on the XTREME-R benchmark?

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