Saved in:
Bibliographic Details
Main Authors: Vergara-Browne, Tomás, Soto, Álvaro, Aizawa, Akiko
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2404.03147
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910494955143168
author Vergara-Browne, Tomás
Soto, Álvaro
Aizawa, Akiko
author_facet Vergara-Browne, Tomás
Soto, Álvaro
Aizawa, Akiko
contents We introduce eigenpruning, a method that removes singular values from weight matrices in an LLM to improve its performance in a particular task. This method is inspired by interpretability methods designed to automatically find subnetworks of a model which solve a specific task. In our tests, the pruned model outperforms the original model by a large margin, while only requiring minimal computation to prune the weight matrices. In the case of a small synthetic task in integer multiplication, the Phi-2 model can improve its accuracy in the test set from 13.75% to 97.50%. Interestingly, these results seem to indicate the existence of a computation path that can solve the task very effectively, but it was not being used by the original model. Finally, we publicly release our implementation.
format Preprint
id arxiv_https___arxiv_org_abs_2404_03147
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Eigenpruning: an Interpretability-Inspired PEFT Method
Vergara-Browne, Tomás
Soto, Álvaro
Aizawa, Akiko
Machine Learning
Artificial Intelligence
We introduce eigenpruning, a method that removes singular values from weight matrices in an LLM to improve its performance in a particular task. This method is inspired by interpretability methods designed to automatically find subnetworks of a model which solve a specific task. In our tests, the pruned model outperforms the original model by a large margin, while only requiring minimal computation to prune the weight matrices. In the case of a small synthetic task in integer multiplication, the Phi-2 model can improve its accuracy in the test set from 13.75% to 97.50%. Interestingly, these results seem to indicate the existence of a computation path that can solve the task very effectively, but it was not being used by the original model. Finally, we publicly release our implementation.
title Eigenpruning: an Interpretability-Inspired PEFT Method
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2404.03147