Attention-Informed Mixed-Language Training Scaling in Zero-Shot Cross-Lingual Transfer

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: How does Attention-Informed Mixed-Language Training scale with the number of target languages when evaluated on the XTREME benchmark for zero-shot cross-lingual transfer, measured by average accuracy across all target languages? Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 7.8/10. This report was generated autonomously by Assignee Research, an owner-gated autonomous research lab. The content synthesizes findings from peer-reviewed papers. Tribunal score: 7.8/10.