Article

Discovering Latent Sycophantic Patterns in LLMs via Distributional Anomaly Detection

2026-05-08

Abstract excerpt

Large language models (LLMs) often exhibit sycophancy—the tendency to generate responses that align with a user’s stated beliefs or desires, even when those responses are objectively incorrect or misleading. While sycophantic behavior is typically studied through explicit prompt perturbations or behavioral benchmarks, latent sycophantic patterns may persist hidden within a model’s internal representations and outp...

Topics

Open a Topic to create a Post that cites this publication.

Identifiers and source

Literature Corpus work: e5b1accf-579b-53a0-b101-ab3864d3141c
DOI: 10.14293/pr2199.003532.v1

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

Discovering Latent Sycophantic Patterns in LLMs via Distributional Anomaly DetectionDOI 10.14293/pr2199.003532.v1Read

Discovering Latent Sycophantic Patterns in LLMs via Distributional Anomaly DetectionSemantic proximity Read

Emergent Numeric Bias in Large Language Models: An Empirical Study on the Anomalous Recurrence of the Number 27 Across Independent SessionsSemantic proximity Read

Finding Rare Racial Stereotypes in LLM Responses via Adversarial Anomaly DetectionSemantic proximity Read

Finding Rare Racial Stereotypes in LLM Responses via Adversarial Anomaly DetectionSemantic proximity Read

Anchoring Bias in Large Language Models: An Experimental StudySemantic proximity Read

Adaptive Contextual Preference Alignment for Large Language ModelsSemantic proximity Read

Optimizing Human-Controlled Preference Alignment in Large Language Models via Dense Token Masking: A Methodological ApproachSemantic proximity Read

Noise-Robust Preference Alignment for Large Language Models via Confidence Estimation and Adaptive OptimizationSemantic proximity Read

Self-Pluralising Culture Alignment for Large Language ModelsSemantic proximity Read

Negative Preference Reduction in Large Language Model Unlearning: An Experimental ApproachSemantic proximity Read

The Unreliable Judges: Assessing Reproducibility and Self-Preference Bias of LLMs as Free-Text EvaluatorsSemantic proximity Read

A Survey of Large Language Models: Evolution, Architectures, Adaptation, Benchmarking, Applications, Challenges, and Societal ImplicationsSemantic proximity Read