Think Again! The Effect of Test-Time Compute on Preferences, Opinions, and Beliefs of Large Language Models

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Kour, George, Nakash, Itay, Anaby-Tavor, Ateret, Shmueli-Scheuer, Michal
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910968588533760
author Kour, George
Nakash, Itay
Anaby-Tavor, Ateret
Shmueli-Scheuer, Michal
author_facet Kour, George
Nakash, Itay
Anaby-Tavor, Ateret
Shmueli-Scheuer, Michal
contents As Large Language Models (LLMs) become deeply integrated into human life and increasingly influence decision-making, it's crucial to evaluate whether and to what extent they exhibit subjective preferences, opinions, and beliefs. These tendencies may stem from biases within the models, which may shape their behavior, influence the advice and recommendations they offer to users, and potentially reinforce certain viewpoints. This paper presents the Preference, Opinion, and Belief survey (POBs), a benchmark developed to assess LLMs' subjective inclinations across societal, cultural, ethical, and personal domains. We applied our benchmark to evaluate leading open- and closed-source LLMs, measuring desired properties such as reliability, neutrality, and consistency. In addition, we investigated the effect of increasing the test-time compute, through reasoning and self-reflection mechanisms, on those metrics. While effective in other tasks, our results show that these mechanisms offer only limited gains in our domain. Furthermore, we reveal that newer model versions are becoming less consistent and more biased toward specific viewpoints, highlighting a blind spot and a concerning trend. POBS: https://ibm.github.io/POBS
format Preprint
id arxiv_https___arxiv_org_abs_2505_19621
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Think Again! The Effect of Test-Time Compute on Preferences, Opinions, and Beliefs of Large Language Models
Kour, George
Nakash, Itay
Anaby-Tavor, Ateret
Shmueli-Scheuer, Michal
Artificial Intelligence
Computation and Language
As Large Language Models (LLMs) become deeply integrated into human life and increasingly influence decision-making, it's crucial to evaluate whether and to what extent they exhibit subjective preferences, opinions, and beliefs. These tendencies may stem from biases within the models, which may shape their behavior, influence the advice and recommendations they offer to users, and potentially reinforce certain viewpoints. This paper presents the Preference, Opinion, and Belief survey (POBs), a benchmark developed to assess LLMs' subjective inclinations across societal, cultural, ethical, and personal domains. We applied our benchmark to evaluate leading open- and closed-source LLMs, measuring desired properties such as reliability, neutrality, and consistency. In addition, we investigated the effect of increasing the test-time compute, through reasoning and self-reflection mechanisms, on those metrics. While effective in other tasks, our results show that these mechanisms offer only limited gains in our domain. Furthermore, we reveal that newer model versions are becoming less consistent and more biased toward specific viewpoints, highlighting a blind spot and a concerning trend. POBS: https://ibm.github.io/POBS
title Think Again! The Effect of Test-Time Compute on Preferences, Opinions, and Beliefs of Large Language Models
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2505.19621