Layer-wise Feature Aggregation Effects on mBERT Zero-shot Cross-lingual Performance in XTREME-R

Multilingual BERT (mBERT), a language model pre-trained on large multilingual corpora, has impressive zero-shot cross-lingual transfer capabilities and performs surprisingly well on zero-shot POS tagging and Named Entity Recognition (NER), as well as on cross-lingual model transfer. At present, the mainstream methods to solve the cross-lingual downstream tasks are always using the last transformer layer's output of mBERT as the representation of linguistic information. In this work, we explore the complementary property of lower layers to the last transformer layer of mBERT. A feature aggregat

Research goal: What is the effect of layer-wise feature aggregation (e.g., concatenating vs. averaging vs. weighted sum of transformer layers) on the zero-shot cross-lingual performance of mBERT on XTREME-R, compared to using only the last layer?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 9.0/10.