Comparison of Intermediate-Task Training and Multilingual Pretraining for Zero-Shot Cross-Lingual Accuracy on XTREME
Description
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 intermediate-task training compare to multilingual pretraining (e.g., mT5, XLM-R) in improving zero-shot cross-lingual accuracy on XTREME, evaluated using task-specific metrics like NER F1 and XNLI accuracy?
Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.5/10.
Notes
Files
paper.pdf
Files
(86.6 kB)
| Name | Size | Download all |
|---|---|---|
|
md5:4094eba796569aae76998efa956a1068
|
86.6 kB | Preview Download |