The More Polypersonal the Better - A Short Look on Space Geometry of Fine-Tuned Layers

P. 13–22.

Sergei Kudriashov, Veronika Zykova, Stepanova A., Jacob Raskind, Eduard Klyshinsky

The interpretation of deep learning models is a rapidly growing field,
with particular interest in language models. There are various approaches to this
task, including training simpler models to replicate neural network predictions and
analyzing the latent space of the model. The latter method allows us to not only
identify patterns in the model’s decision-making process, but also understand the
features of its internal structure. In this paper, we analyze the changes in the internal
representation of the BERT model when it is trained with additional grammatical
modules and data containing new grammatical features (polypersonality). We find
that adding even a single grammatical layer causes the model to separate the new
and old grammatical systems within itself, improving the overall performance on
perplexity metrics.

Keywords: Topological data analysis, BERT, latent space, Interpretation of the Language Model

Publication based on the results of: The expression of stance in L1 and L2 learners’ academic texts (2024)

In book

Advances in Neural Computation, Machine Learning, and Cognitive Research VIII, Selected Papers from the XXVI International Conference on Neuroinformatics, October 21-25, 2024, Moscow, Russia

Vol. VIII. , Cham: Springer, 2024.