paper-with-me

Papers

Sycophantic Anchors: Localizing and Quantifying User Agreement in Reasoning Models

2026-01-29 · Jacek Duszenko arxiv

Reasoning models frequently agree with incorrect user suggestions -- a behavior known as sycophancy. However, it is unclear where in the reasoning trace this agreement originates and how strong the commitment is. We introduce \emph{sycophantic anchors} -- sentences identified via counterfactual analysis that commit models to user agreement. Across four reasoning models spanning three architecture families (Llama, Qwen, Falcon-hybrid) and 1.5B--8B parameters, we analyze over 200,000 counterfactual rollouts and show that linear probes reliably detect sycophantic anchors (74--85\% balanced accuracy), outperforming text-only baselines at high commitment levels -- confirming they capture internal states beyond surface vocabulary. Regressors further predict commitment strength from activations ($R^2$ up to 0.74). We observe a consistent asymmetry: sycophancy leaves a stronger mechanistic footprint than correct reasoning. We also find that sycophancy builds gradually during generation rather than being determined by the prompt. These findings enable sentence-level detection and quantification of model misalignment mid-inference.

📄 PDF Abstract BibTeX arXiv:2601.21183

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Sycophancy Is Not One Thing: Causal Separation of Sycophantic Behaviors in LLMs

2025-09-25 · Daniel Vennemeyer, Phan Anh Duong, Tiffany Zhan, Tianyu Jiang arxiv

Large language models (LLMs) often exhibit sycophantic behaviors -- such as excessive agreement with or flattery of the user -- but it is unclear whether these behaviors arise from a single mechanism or multiple distinct…

Sycophantic Praise: Evaluating Excessive Praise in Language Models

2026-06-05 · Daniel Vennemeyer, Phan Anh Duong, Meryl Ye, Ruihong Huang 외 arxiv

Sycophancy in language models is typically studied as excessive agreement or validation, while explicit praise and flattery have received comparatively little attention. We argue that sycophantic praise is a distinct ali…

Complacent, Not Sycophantic: Reframing Large Language Models and Designing AI Literacy for Complacent Machines

2026-05-14 · Federico Germani, Giovanni Spitale arxiv

Large language models are often described as sycophantic, in the sense that they appear to flatter users or mirror their beliefs. We argue that this label is conceptually misleading: sycophancy implies motives and strate…

Linear Probe Penalties Reduce LLM Sycophancy

2024-12-01 · Henry Papadatos, Rachel Freedman

Large language models (LLMs) are often sycophantic, prioritizing agreement with their users over accurate or objective statements. This problematic behavior becomes more pronounced during reinforcement learning from huma…

From Sycophancy to Sensemaking: Premise Governance for Human-AI Decision Making

2026-02-02 · Raunak Jain arxiv

As LLMs expand from assistance to decision support, a dangerous pattern emerges: fluent agreement without calibrated judgment. Low-friction assistants can become sycophantic, baking in implicit assumptions and pushing ve…

Answer GenerationDecision Making