Who Endorsed It? Measuring Authority Bias Across Expertise Levels in Language Models
On the map: LLM reasoning → Reasoning benchmarks and evaluation
Abstract (arXiv)
Prior research demonstrates that performance of language models on reasoning tasks can be influenced by suggestions, hints and endorsements. However, the influence of endorsement source credibility remains underexplored. We investigate whether language models exhibit systematic bias based on the perceived expertise of the provider of the endorsement. Across 4 datasets spanning mathematical, legal, and medical reasoning, we evaluate 11 models using personas representing four expertise levels per domain. Our results reveal that models are increasingly susceptible to incorrect/misleading endorsements as source expertise increases, with higher-authority sources inducing not only accuracy degradation but also increased confidence in wrong answers. We also show that this authority bias is mechanistically encoded within the model and a model can be steered away from the bias, thereby improving its performance even when an expert gives a misleading endorsement.
Key facts. Cited by 7 papers in the map (data through 7 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 13 of its 27 references that are in the map. First posted on arXiv 19 Jan 2026.
Builds on
13 of 27 references.
Closest ancestors · 2
Same direction, newest first
- 2025Assessing Judging Bias in Large Reasoning Models: An Empirical Study14 Apr
- 2024From Calculation to Adjudication: Examining LLM judges on Mathematical Reasoning Tasks6 Sep
Other references · 11
- 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
- 2020What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Examscited by 1,324
- 2023Large Language Models are not Fair Evaluatorscited by 1,050
- 2024Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluatorscited by 865
- 2023Steering Language Models With Activation Engineeringcited by 824
- 2017Program Induction by Rationale Generation : Learning to Solve and Explain Algebraic Word Problemscited by 642
- 2024Humans or LLMs as the Judge? A Study on Judgement Biasescited by 314
- 2024Judging the Judges: A Systematic Study of Position Bias in LLM-as-a-Judgecited by 249
- 2023Verbosity Bias in Preference Labeling by Large Language Modelscited by 89
- 2025LEXam: Benchmarking Legal Reasoning on 340 Law Examscited by 52
- 2025Enhancing Jailbreak Attacks on LLMs via Persona Promptscited by 10
Built on by
Citing papers per year: 2026: 7.
7 papers.
- 2026A Mechanistic View of Authority Hierarchy in LLM Sycophancycited by 6
- 2026Fragile Thoughts: How Large Language Models Handle Chain-of-Thought Perturbationscited by 5
- 2026The Authority Expectancy Effect in Multi-Party Conflictcited by 0
- 2026How Do Language Models Choose Between Context and Memory?cited by 0
- 2026GradeTrap: Authority Cues in Images Shift VLM Judgments Despite Explicit Instructions to Ignore Themcited by 0
- 2026Silence Is Endorsement: Verification-Status Laundering in LLM Agent Pipelinescited by 0
- 2026Authority Bias in Language Models: Source Deference and User Agreement Are Not Interchangeablecited by 0
- 2026Authority Bias in Language Models: Source Deference and User Agreement Are Not Interchangeable29 Sep 2026
- 2026GradeTrap: Authority Cues in Images Shift VLM Judgments Despite Explicit Instructions to Ignore Them5 Sep 2026
- 2026How Do Language Models Choose Between Context and Memory?1 Sep 2026
- 2026The Authority Expectancy Effect in Multi-Party Conflict8 Aug 2026
- 2026Silence Is Endorsement: Verification-Status Laundering in LLM Agent Pipelines25 Jul 2026
- 2026A Mechanistic View of Authority Hierarchy in LLM Sycophancy1 Jul 2026
- 2026Fragile Thoughts: How Large Language Models Handle Chain-of-Thought Perturbations11 Feb 2026
Similar, not cited
Close in wording, no citation either way
- 2026Why Expert Alignment Is Hard: Evidence from Subjective Evaluation
- 2025Towards Evaluting Fake Reasoning Bias in Language Models
- 2026Who Do Language Models Think Is Competent? A Mechanistic Analysis of Occupational Bias
- 2026Failing to Falsify: Evaluating and Mitigating Confirmation Bias in Language Models
- 2026Evaluating the False Trust Engendered by LLM Explanations
- 2025The simulation of judgment in LLMs
- 2026Seeing the Reasoning: How LLM Rationales Influence User Trust and Decision-Making in Factual Verification Tasks
- 2025Safer or Luckier? LLMs as Safety Evaluators Are Not Robust to Artifacts