Impact of Model Architecture on Inference Latency and Accuracy in Zero-Shot Cross-Lingual Transfer

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 impact of model architecture (e.g., Transformer vs. Mixture-of-Experts) on inference latency and accuracy when fine-tuning on English versus multilingual intermediate tasks for zero-shot cross-lingual transfer on XTREME-R?

Autonomous synthesis report generated by Assignee Research. Tribunal consensus score: 7.7/10.