GrounDial: Human-norm Grounded Safe Dialog Response Generation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Kim, Siwon, Dai, Shuyang, Kachuee, Mohammad, Ray, Shayan, Taghavi, Tara, Yoon, Sungroh
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911776887537664
author Kim, Siwon
Dai, Shuyang
Kachuee, Mohammad
Ray, Shayan
Taghavi, Tara
Yoon, Sungroh
author_facet Kim, Siwon
Dai, Shuyang
Kachuee, Mohammad
Ray, Shayan
Taghavi, Tara
Yoon, Sungroh
contents Current conversational AI systems based on large language models (LLMs) are known to generate unsafe responses, agreeing to offensive user input or including toxic content. Previous research aimed to alleviate the toxicity, by fine-tuning LLM with manually annotated safe dialogue histories. However, the dependency on additional tuning requires substantial costs. To remove the dependency, we propose GrounDial, where response safety is achieved by grounding responses to commonsense social rules without requiring fine-tuning. A hybrid approach of in-context learning and human-norm-guided decoding of GrounDial enables the response to be quantitatively and qualitatively safer even without additional data or tuning.
format Preprint
id arxiv_https___arxiv_org_abs_2402_08968
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle GrounDial: Human-norm Grounded Safe Dialog Response Generation
Kim, Siwon
Dai, Shuyang
Kachuee, Mohammad
Ray, Shayan
Taghavi, Tara
Yoon, Sungroh
Artificial Intelligence
Current conversational AI systems based on large language models (LLMs) are known to generate unsafe responses, agreeing to offensive user input or including toxic content. Previous research aimed to alleviate the toxicity, by fine-tuning LLM with manually annotated safe dialogue histories. However, the dependency on additional tuning requires substantial costs. To remove the dependency, we propose GrounDial, where response safety is achieved by grounding responses to commonsense social rules without requiring fine-tuning. A hybrid approach of in-context learning and human-norm-guided decoding of GrounDial enables the response to be quantitatively and qualitatively safer even without additional data or tuning.
title GrounDial: Human-norm Grounded Safe Dialog Response Generation
topic Artificial Intelligence
url https://arxiv.org/abs/2402.08968