Saved in:
Bibliographic Details
Main Authors: Hawkins, Will, Mittelstadt, Brent, Russell, Chris
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2410.15821
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913557442985984
author Hawkins, Will
Mittelstadt, Brent
Russell, Chris
author_facet Hawkins, Will
Mittelstadt, Brent
Russell, Chris
contents Fine-tuning language models has become increasingly popular following the proliferation of open models and improvements in cost-effective parameter efficient fine-tuning. However, fine-tuning can influence model properties such as safety. We assess how fine-tuning can impact different open models' propensity to output toxic content. We assess the impacts of fine-tuning Gemma, Llama, and Phi models on toxicity through three experiments. We compare how toxicity is reduced by model developers during instruction-tuning. We show that small amounts of parameter-efficient fine-tuning on developer-tuned models via low-rank adaptation on a non-adversarial dataset can significantly alter these results across models. Finally, we highlight the impact of this in the wild, demonstrating how toxicity rates of models fine-tuned by community contributors can deviate in hard-to-predict ways.
format Preprint
id arxiv_https___arxiv_org_abs_2410_15821
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle The effect of fine-tuning on language model toxicity
Hawkins, Will
Mittelstadt, Brent
Russell, Chris
Artificial Intelligence
Fine-tuning language models has become increasingly popular following the proliferation of open models and improvements in cost-effective parameter efficient fine-tuning. However, fine-tuning can influence model properties such as safety. We assess how fine-tuning can impact different open models' propensity to output toxic content. We assess the impacts of fine-tuning Gemma, Llama, and Phi models on toxicity through three experiments. We compare how toxicity is reduced by model developers during instruction-tuning. We show that small amounts of parameter-efficient fine-tuning on developer-tuned models via low-rank adaptation on a non-adversarial dataset can significantly alter these results across models. Finally, we highlight the impact of this in the wild, demonstrating how toxicity rates of models fine-tuned by community contributors can deviate in hard-to-predict ways.
title The effect of fine-tuning on language model toxicity
topic Artificial Intelligence
url https://arxiv.org/abs/2410.15821