Sycophantic response patterns in Large Language Models (LLMs) have been increasingly claimed in the literature. We review methodological challenges in measuring LLM sycophancy and identify five core operationalizations. Despite sycophancy being inherently human-centric, current research does not evaluate human perception. Our analysis highlights the difficulties in distinguishing sycophantic responses from related concepts in AI alignment and offers actionable recommendations for future research.
@article{arxiv.2512.00656,
title = {Sycophancy Claims about Language Models: The Missing Human-in-the-Loop},
author = {Jan Batzner and Volker Stocker and Stefan Schmid and Gjergji Kasneci},
journal= {arXiv preprint arXiv:2512.00656},
year = {2025}
}
Comments
NeurIPS 2025 Workshop on LLM Evaluation and ICLR 2025 Workshop on Bi-Directional Human-AI Alignment