Attribute Controlled Fine-tuning for Large Language Models: A Case Study on Detoxification

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Meng, Tao, Mehrabi, Ninareh, Goyal, Palash, Ramakrishna, Anil, Galstyan, Aram, Zemel, Richard, Chang, Kai-Wei, Gupta, Rahul, Peris, Charith
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866913535976538112
author Meng, Tao
Mehrabi, Ninareh
Goyal, Palash
Ramakrishna, Anil
Galstyan, Aram
Zemel, Richard
Chang, Kai-Wei
Gupta, Rahul
Peris, Charith
author_facet Meng, Tao
Mehrabi, Ninareh
Goyal, Palash
Ramakrishna, Anil
Galstyan, Aram
Zemel, Richard
Chang, Kai-Wei
Gupta, Rahul
Peris, Charith
contents We propose a constraint learning schema for fine-tuning Large Language Models (LLMs) with attribute control. Given a training corpus and control criteria formulated as a sequence-level constraint on model outputs, our method fine-tunes the LLM on the training corpus while enhancing constraint satisfaction with minimal impact on its utility and generation quality. Specifically, our approach regularizes the LLM training by penalizing the KL divergence between the desired output distribution, which satisfies the constraints, and the LLM's posterior. This regularization term can be approximated by an auxiliary model trained to decompose the sequence-level constraints into token-level guidance, allowing the term to be measured by a closed-form formulation. To further improve efficiency, we design a parallel scheme for concurrently updating both the LLM and the auxiliary model. We evaluate the empirical performance of our approach by controlling the toxicity when training an LLM. We show that our approach leads to an LLM that produces fewer inappropriate responses while achieving competitive performance on benchmarks and a toxicity detection task.
format Preprint
id arxiv_https___arxiv_org_abs_2410_05559
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Attribute Controlled Fine-tuning for Large Language Models: A Case Study on Detoxification
Meng, Tao
Mehrabi, Ninareh
Goyal, Palash
Ramakrishna, Anil
Galstyan, Aram
Zemel, Richard
Chang, Kai-Wei
Gupta, Rahul
Peris, Charith
Computation and Language
We propose a constraint learning schema for fine-tuning Large Language Models (LLMs) with attribute control. Given a training corpus and control criteria formulated as a sequence-level constraint on model outputs, our method fine-tunes the LLM on the training corpus while enhancing constraint satisfaction with minimal impact on its utility and generation quality. Specifically, our approach regularizes the LLM training by penalizing the KL divergence between the desired output distribution, which satisfies the constraints, and the LLM's posterior. This regularization term can be approximated by an auxiliary model trained to decompose the sequence-level constraints into token-level guidance, allowing the term to be measured by a closed-form formulation. To further improve efficiency, we design a parallel scheme for concurrently updating both the LLM and the auxiliary model. We evaluate the empirical performance of our approach by controlling the toxicity when training an LLM. We show that our approach leads to an LLM that produces fewer inappropriate responses while achieving competitive performance on benchmarks and a toxicity detection task.
title Attribute Controlled Fine-tuning for Large Language Models: A Case Study on Detoxification
topic Computation and Language
url https://arxiv.org/abs/2410.05559