Think Again! The Effect of Test-Time Compute on Preferences, Opinions, and Beliefs of Large Language Models
On the map: LLM reasoning → Reasoning benchmarks and evaluation
Abstract (arXiv)
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
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 19 of its 44 references that are in the map. First posted on arXiv 26 May 2025.
Builds on
19 of 44 references.
Closest ancestors · 3
Same direction, newest first
- 2025Does Reasoning Introduce Bias? A Study of Social Bias Evaluation and Mitigation in LLM Reasoning21 Feb
- 2024DeepSeek-V3 Technical Report27 Dec
- 2022Towards Reasoning in Large Language Models: A Survey20 Dec
Other references · 16
- 2023GPT-4 Technical Reportcited by 18,132
- 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
- 2024Qwen2.5 Technical Reportcited by 4,713
- 2024GPT-4o System Cardcited by 4,467
- 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,928
- 2024DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Modelcited by 1,093
- 2023Towards Measuring the Representation of Subjective Global Opinions in Language Modelscited by 424
- 2024Large Language Models for Mathematical Reasoning: Progresses and Challengescited by 283
- 2024Forest-of-Thought: Scaling Test-Time Compute for Enhancing LLM Reasoningcited by 110
- 2024Assessing Political Bias in Large Language Modelscited by 32
- 2019Fair is Better than Sensational:Man is to Doctor as Woman is to Doctorcited by 27
- 2024The Colorful Future of LLMs: Evaluating and Improving LLMs as Emotional Supporters for Queer Youthcited by 20
4 more
- 2024The LLM Effect: Are Humans Truly Using LLMs, or Are They Being Influenced By Them Instead?cited by 14
- 2023Being Right for Whose Right Reasons?cited by 12
- 2024Whose Side Are You On? Investigating the Political Stance of Large Language Modelscited by 10
- 2025Improving Neutral Point-of-View Generation with Data- and Parameter-Efficient RLcited by 3
Built on by
Citing papers per year: 2026: 2.
2 papers.
Similar, not cited
Close in wording, no citation either way
- 2023Can LLMs Capture Human Preferences?
- 2024Dissecting Human and LLM Preferences
- 2025Rethinking Human Preference Evaluation of LLM Rationales
- 2025Alignment Revisited: Are Large Language Models Consistent in Stated and Revealed Preferences?
- 2026Evaluating LLM-Generated Preference Distributions
- 2026Do LLMs Have Values? A Quantitative Analysis and Alignment Framework for Values in Large Language Models
- 2025Do Language Models Think Consistently? A Study of Value Preferences Across Varying Response Lengths
- 2025Beyond the Surface: Measuring Self-Preference in LLM Judgments