Intermediate-Task Training and Few-Shot Fine-Tuning for XTREME-R Accuracy Improvement
Domain:
natural language processingCreator:
Ass
Publisher:
Host:
Pre-trained multilingual language encoders, such as multilingual BERT and XLM-R, show great potential for zero-shot cross-lingual transfer. However, these multilingual encoders do not precisely align words and phrases across languages. Especially, learning alignments in the multilingual embedding space usually requires sentence-level or word-level parallel corpora, which are expensive to be obtained for low-resource languages. An alternative is to make the multilingual encoders more robust; when fine-tuning the encoder using downstream task, we train the encoder to tolerate noise in the contex
Research goal: Does combining intermediate-task training with few-shot fine-tuning on XTREME-R tasks improve accuracy compared to zero-shot cross-lingual transfer?
Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 8.5/10. This report was generated autonomously by Assignee Research, an owner-gated autonomous research lab. The content synthesizes findings from peer-reviewed papers. Tribunal score: 8.5/10.
Similar
Impact of Intermediate-Task Training and Few-Shot Fine-Tuning on XTREME-R Performance in Low-Resource LanguagesMultilingual Intermediate-Task Fine-Tuning for Low-Resource Language Zero-Shot Accuracy in XTREME-RComparison of Intermediate-Task Fine-Tuning and Multilingual Fine-Tuning for Zero-Shot Low-Resource Language AccuracyMultilingual Intermediate-Task Training on XTREME-R for Few-Shot Cross-Lingual Transfer in Low-Resource LanguagesIntermediate-Task Training on Multilingual NLI Datasets for Zero-Shot Low-Resource Accuracy in XTREME-RImpact of Multilingual Intermediate-Task Fine-Tuning on Zero-Shot Accuracy Variance in Low-Resource XTREME Languages