Not All Models Localize Linguistic Knowledge in the Same Place: A Layer-wise Probing on BERToids' Representations

by   Mohsen Fayyaz, et al.

Most of the recent works on probing representations have focused on BERT, with the presumption that the findings might be similar to the other models. In this work, we extend the probing studies to two other models in the family, namely ELECTRA and XLNet, showing that variations in the pre-training objectives or architectural choices can result in different behaviors in encoding linguistic information in the representations. Most notably, we observe that ELECTRA tends to encode linguistic knowledge in the deeper layers, whereas XLNet instead concentrates that in the earlier layers. Also, the former model undergoes a slight change during fine-tuning, whereas the latter experiences significant adjustments. Moreover, we show that drawing conclusions based on the weight mixing evaluation strategy – which is widely used in the context of layer-wise probing – can be misleading given the norm disparity of the representations across different layers. Instead, we adopt an alternative information-theoretic probing with minimum description length, which has recently been proven to provide more reliable and informative results.


page 6

page 7

page 12

page 13


How transfer learning impacts linguistic knowledge in deep NLP models?

Transfer learning from pre-trained neural language models towards downst...

What Happens To BERT Embeddings During Fine-tuning?

While there has been much recent work studying how linguistic informatio...

Linguistic Profiling of a Neural Language Model

In this paper we investigate the linguistic knowledge learned by a Neura...

Revisiting Few-sample BERT Fine-tuning

We study the problem of few-sample fine-tuning of BERT contextual repres...

Information-Theoretic Probing with Minimum Description Length

To measure how well pretrained representations encode some linguistic pr...

EigenNoise: A Contrastive Prior to Warm-Start Representations

In this work, we present a naive initialization scheme for word vectors ...

A Matter of Framing: The Impact of Linguistic Formalism on Probing Results

Deep pre-trained contextualized encoders like BERT (Delvin et al., 2019)...