Acquiring Bidirectionality via Large and Small Language Models
This addresses a bottleneck in token-classification tasks for NLP practitioners, offering a method to enhance performance, especially in rare domains and few-shot settings, though it is incremental.
The paper tackled the problem of unidirectional large language models underperforming in token-classification tasks due to lack of bidirectionality, and by training a small backward LM and concatenating its representations, they improved benchmark performance by over 10 points in named entity recognition.
Using token representation from bidirectional language models (LMs) such as BERT is still a widely used approach for token-classification tasks. Even though there exist much larger unidirectional LMs such as Llama-2, they are rarely used to replace the token representation of bidirectional LMs. In this work, we hypothesize that their lack of bidirectionality is keeping them behind. To that end, we propose to newly train a small backward LM and concatenate its representations to those of existing LM for downstream tasks. Through experiments in named entity recognition, we demonstrate that introducing backward model improves the benchmark performance more than 10 points. Furthermore, we show that the proposed method is especially effective for rare domains and in few-shot learning settings.