Saved in:
Bibliographic Details
Main Authors: Klein, Tassilo, Nabi, Moin
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2401.08491
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908385627078656
author Klein, Tassilo
Nabi, Moin
author_facet Klein, Tassilo
Nabi, Moin
contents The generation of toxic content by large language models (LLMs) remains a critical challenge for the safe deployment of language technology. We propose a novel framework for implicit knowledge editing and controlled text generation by fine-tuning LLMs with a prototype-based contrastive perplexity objective. Central to our method is the construction of hard negatives - toxic outputs that are generated through adversarial paraphrasing to be semantically similar and model probability to their non-toxic counterparts. By training on these challenging and realistic pairs, our approach ensures robust and stable contrastive optimization. Experimental results in the domain of detoxification demonstrate that our method significantly reduces toxic generation while maintaining strong performance on downstream tasks such as commonsense reasoning and reading comprehension. Our findings highlight the effectiveness of exploiting hard negatives for attribute-aware fine-tuning.
format Preprint
id arxiv_https___arxiv_org_abs_2401_08491
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Contrastive Perplexity for Controlled Generation: An Application in Detoxifying Large Language Models
Klein, Tassilo
Nabi, Moin
Computation and Language
Machine Learning
The generation of toxic content by large language models (LLMs) remains a critical challenge for the safe deployment of language technology. We propose a novel framework for implicit knowledge editing and controlled text generation by fine-tuning LLMs with a prototype-based contrastive perplexity objective. Central to our method is the construction of hard negatives - toxic outputs that are generated through adversarial paraphrasing to be semantically similar and model probability to their non-toxic counterparts. By training on these challenging and realistic pairs, our approach ensures robust and stable contrastive optimization. Experimental results in the domain of detoxification demonstrate that our method significantly reduces toxic generation while maintaining strong performance on downstream tasks such as commonsense reasoning and reading comprehension. Our findings highlight the effectiveness of exploiting hard negatives for attribute-aware fine-tuning.
title Contrastive Perplexity for Controlled Generation: An Application in Detoxifying Large Language Models
topic Computation and Language
Machine Learning
url https://arxiv.org/abs/2401.08491