Cross-Lingual Transfer of Intermediate-Task Training in 30B Models on XTREME-R Across Language Families

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 effectiveness of intermediate-task training on English tasks translate to consistent improvements in downstream task accuracy across different language families (e.g., Indo-European vs. Sino-Tibetan) when evaluated on the XTREME-R benchmark with 30B parameter models?

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