Saved in:
Bibliographic Details
Main Authors: Zavoral, Patrik, Variš, Dušan, Bojar, Ondřej
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2410.13802
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917806988066816
author Zavoral, Patrik
Variš, Dušan
Bojar, Ondřej
author_facet Zavoral, Patrik
Variš, Dušan
Bojar, Ondřej
contents The Transformer model has a tendency to overfit various aspects of the training data, such as the overall sequence length. We study elementary string edit functions using a defined set of error indicators to interpret the behaviour of the sequence-to-sequence Transformer. We show that generalization to shorter sequences is often possible, but confirm that longer sequences are highly problematic, although partially correct answers are often obtained. Additionally, we find that other structural characteristics of the sequences, such as subsegment length, may be equally important. We hypothesize that the models learn algorithmic aspects of the tasks simultaneously with structural aspects but adhering to the structural aspects is unfortunately often preferred by Transformer when they come into conflict.
format Preprint
id arxiv_https___arxiv_org_abs_2410_13802
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Adversarial Testing as a Tool for Interpretability: Length-based Overfitting of Elementary Functions in Transformers
Zavoral, Patrik
Variš, Dušan
Bojar, Ondřej
Machine Learning
The Transformer model has a tendency to overfit various aspects of the training data, such as the overall sequence length. We study elementary string edit functions using a defined set of error indicators to interpret the behaviour of the sequence-to-sequence Transformer. We show that generalization to shorter sequences is often possible, but confirm that longer sequences are highly problematic, although partially correct answers are often obtained. Additionally, we find that other structural characteristics of the sequences, such as subsegment length, may be equally important. We hypothesize that the models learn algorithmic aspects of the tasks simultaneously with structural aspects but adhering to the structural aspects is unfortunately often preferred by Transformer when they come into conflict.
title Adversarial Testing as a Tool for Interpretability: Length-based Overfitting of Elementary Functions in Transformers
topic Machine Learning
url https://arxiv.org/abs/2410.13802