Saved in:
Bibliographic Details
Main Authors: Rennard, Virgile, Xypolopoulos, Christos, Vazirgiannis, Michalis
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2410.13517
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909377570537472
author Rennard, Virgile
Xypolopoulos, Christos
Vazirgiannis, Michalis
author_facet Rennard, Virgile
Xypolopoulos, Christos
Vazirgiannis, Michalis
contents Large language models (LLMs) inherit biases from their training data and alignment processes, influencing their responses in subtle ways. While many studies have examined these biases, little work has explored their robustness during interactions. In this paper, we introduce a novel approach where two instances of an LLM engage in self-debate, arguing opposing viewpoints to persuade a neutral version of the model. Through this, we evaluate how firmly biases hold and whether models are susceptible to reinforcing misinformation or shifting to harmful viewpoints. Our experiments span multiple LLMs of varying sizes, origins, and languages, providing deeper insights into bias persistence and flexibility across linguistic and cultural contexts.
format Preprint
id arxiv_https___arxiv_org_abs_2410_13517
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Bias in the Mirror: Are LLMs opinions robust to their own adversarial attacks ?
Rennard, Virgile
Xypolopoulos, Christos
Vazirgiannis, Michalis
Computation and Language
Artificial Intelligence
Large language models (LLMs) inherit biases from their training data and alignment processes, influencing their responses in subtle ways. While many studies have examined these biases, little work has explored their robustness during interactions. In this paper, we introduce a novel approach where two instances of an LLM engage in self-debate, arguing opposing viewpoints to persuade a neutral version of the model. Through this, we evaluate how firmly biases hold and whether models are susceptible to reinforcing misinformation or shifting to harmful viewpoints. Our experiments span multiple LLMs of varying sizes, origins, and languages, providing deeper insights into bias persistence and flexibility across linguistic and cultural contexts.
title Bias in the Mirror: Are LLMs opinions robust to their own adversarial attacks ?
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2410.13517