Can We Use Probing to Better Understand Fine-tuning and Knowledge Distillation of the BERT NLU?

TL;DR

This paper investigates the use of probing methods to understand how fine-tuning and knowledge distillation affect BERT-based NLU models, revealing limitations in current probing techniques for practical insights.

Contribution

The study critically evaluates the effectiveness of probing methods in analyzing BERT fine-tuning and distillation, highlighting the need for decodability metrics.

Findings

Probing methods in their current form are not well suited for practical analysis.

Structural, Edge, and Conditional probes do not account for decoding ease.

Quantification of information decodability is essential for practical applications.

Abstract

In this article, we use probing to investigate phenomena that occur during fine-tuning and knowledge distillation of a BERT-based natural language understanding (NLU) model. Our ultimate purpose was to use probing to better understand practical production problems and consequently to build better NLU models. We designed experiments to see how fine-tuning changes the linguistic capabilities of BERT, what the optimal size of the fine-tuning dataset is, and what amount of information is contained in a distilled NLU based on a tiny Transformer. The results of the experiments show that the probing paradigm in its current form is not well suited to answer such questions. Structural, Edge and Conditional probes do not take into account how easy it is to decode probed information. Consequently, we conclude that quantification of information decodability is critical for many practical applications of the probing paradigm.