Analyzing Individual Neurons in Pre-trained Language Models
This provides insights into the interpretability of neural networks for NLP researchers, though it is incremental as it extends existing analysis from representations to neurons.
The paper tackled the problem of understanding whether individual neurons in pre-trained language models capture linguistic information, finding that small subsets of neurons can predict tasks like morphology and syntax, with lower-level tasks localized in fewer neurons and differences in distribution across architectures such as XLNet being more localized than BERT.
While a lot of analysis has been carried to demonstrate linguistic knowledge captured by the representations learned within deep NLP models, very little attention has been paid towards individual neurons.We carry outa neuron-level analysis using core linguistic tasks of predicting morphology, syntax and semantics, on pre-trained language models, with questions like: i) do individual neurons in pre-trained models capture linguistic information? ii) which parts of the network learn more about certain linguistic phenomena? iii) how distributed or focused is the information? and iv) how do various architectures differ in learning these properties? We found small subsets of neurons to predict linguistic tasks, with lower level tasks (such as morphology) localized in fewer neurons, compared to higher level task of predicting syntax. Our study also reveals interesting cross architectural comparisons. For example, we found neurons in XLNet to be more localized and disjoint when predicting properties compared to BERT and others, where they are more distributed and coupled.