# A Mechanistic View of Authority Hierarchy in LLM Sycophancy

arXiv: https://arxiv.org/abs/2607.00415 · HTML: https://mlsanity.com/paper/2607.00415/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 21 references that are in the map.
- First posted on arXiv 1 Jul 2026.

## Builds on (most cited first, 17 of 17)

- [Understanding intermediate layers using linear classifier probes](https://mlsanity.com/paper/1610.01644/) (2016)
- [Representation Engineering: A Top-Down Approach to AI Transparency](https://mlsanity.com/paper/2310.01405/) (2023)
- [Discovering Language Model Behaviors with Model-Written Evaluations](https://mlsanity.com/paper/2212.09251/) (2022)
- [The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasets](https://mlsanity.com/paper/2310.06824/) (2023)
- [Persona Vectors: Monitoring and Controlling Character Traits in Language Models](https://mlsanity.com/paper/2507.21509/) (2025)
- [Simple synthetic data reduces sycophancy in large language models](https://mlsanity.com/paper/2308.03958/) (2023)
- [When Truth Is Overridden: Uncovering the Internal Origins of Sycophancy in Large Language Models](https://mlsanity.com/paper/2508.02087/) (2025)
- [Assessing Judging Bias in Large Reasoning Models: An Empirical Study](https://mlsanity.com/paper/2504.09946/) (2025)
- [EFUF: Efficient Fine-grained Unlearning Framework for Mitigating Hallucinations in Multimodal Large Language Models](https://mlsanity.com/paper/2402.09801/) (2024)
- [Ask don't tell: Reducing sycophancy in large language models](https://mlsanity.com/paper/2602.23971/) (2026)
- [Self-Steering Language Models](https://mlsanity.com/paper/2504.07081/) (2025)
- [Sycophancy Mitigation Through Reinforcement Learning with Uncertainty-Aware Adaptive Reasoning Trajectories](https://mlsanity.com/paper/2509.16742/) (2025)
- [Who Endorsed It? Measuring Authority Bias Across Expertise Levels in Language Models](https://mlsanity.com/paper/2601.13433/) (2026)
- [Enhancing Reasoning Abilities of Small LLMs with Cognitive Alignment](https://mlsanity.com/paper/2504.09802/) (2025)
- [Belief in Authority: Impact of Authority in Multi-Agent Evaluation Framework](https://mlsanity.com/paper/2601.04790/) (2026)
- [Dissecting Persona-Driven Reasoning in Language Models via Activation Patching](https://mlsanity.com/paper/2507.20936/) (2025)
- [Beyond Superficial Unlearning: Sharpness-Aware Robust Erasure of Hallucinations in Multimodal LLMs](https://mlsanity.com/paper/2601.16527/) (2026)

## Built on by (most cited first, 9 of 9)

- [Authority Bias in Language Models: Source Deference and User Agreement Are Not Interchangeable](https://mlsanity.com/paper/2609.37616/) (2026)
- [GradeTrap: Authority Cues in Images Shift VLM Judgments Despite Explicit Instructions to Ignore Them](https://mlsanity.com/paper/2609.06058/) (2026)
- [Silence Is Endorsement: Verification-Status Laundering in LLM Agent Pipelines](https://mlsanity.com/paper/2609.20211/) (2026)
- [The Authority Expectancy Effect in Multi-Party Conflict](https://mlsanity.com/paper/2608.08026/) (2026)
- [How Do Language Models Choose Between Context and Memory?](https://mlsanity.com/paper/2609.00753/) (2026)
- [Tracing mechanisms of sycophantic agreement in language models](https://mlsanity.com/paper/2609.35822/) (2026)
- [Silent Dissent: LLM Agents That Yield to the Majority Still Represent Their Original Premise](https://mlsanity.com/paper/2610.02702/) (2026)
- [Beyond the Sycophancy Score: How Task, Model, and Pressure Shape LLM Yielding](https://mlsanity.com/paper/2610.08840/) (2026)
- [Claim Framing as a Driver of LLM Sycophancy](https://mlsanity.com/paper/2610.18159/) (2026)

## Abstract (arXiv)

Authority bias poses a critical safety concern in language models: models systematically prioritize social cues from authority figures over factual consistency, swaying their answers based on source credibility rather than evidence. We mechanistically investigate this phenomenon using a controlled medical QA setting, where hints suggesting incorrect answers are attributed to personas of varying expertise. Across Llama-3.1-8B, Qwen3-8B, and Gemma-2-9B, we find that models respond in a graded manner proportional to perceived authority, a hierarchy that is never explicitly prompted but emerges from training. Logit lens analysis and linear/non-linear probing localize this effect to a critical late layer where correct answer representations are actively erased, an erasure that scales with authority level, resists mean vector intervention, and is only partially reversible through chain-of-thought reasoning. Our findings suggest that authority-induced sycophancy is not a surface-level output bias but mechanistic knowledge erasure, a precise, layer-localized overwriting of correct internal representations by high-status authority signals.