Detecting and Controlling Sycophancy with Cascading Linear Features
Summary
Presents an iterative data generation pipeline to isolate cascading linear features responsible for sycophancy in language models, enabling detection, scoring, and steering with lower computational cost than baselines.
View Cached Full Text
Cached at: 06/26/26, 05:11 AM
# Detecting and Controlling Sycophancy with Cascading Linear Features Source: [https://arxiv.org/abs/2606.26155](https://arxiv.org/abs/2606.26155) [View PDF](https://arxiv.org/pdf/2606.26155) > Abstract:Interpreting and controlling model behaviors through activation steering methods requires many pairs of contrastive samples that clearly exhibit desired or undesired behavior\. These data pairs determine the degree to which interpretability frameworks can reliably detect model features responsible for a behavior, and therefore the ability to steer models toward or away from such behavior\. In this work, we present an iterative data generation pipeline that isolates cascading linear features responsible for a behavior\. Specifically, we show how moving beyond simple binary pairs of samples, and instead isolating samples that show degrees of features that scale linearly with behavior, allows for better disentanglement of features\. We focus on detecting and steering away from sycophancy \-\- the tendency of language models to prioritize user validation\. We demonstrate that sycophancy features discovered through cascading samples form linearly separable subspaces, and allow for selection of model activations that more clearly correspond to the desired behavior than baseline approaches\. We also evaluate their ability to enable detection, deterministic scoring, and robust steering, and see that they either match or outperform LLM\-as\-a\-judge and system prompting baselines while providing lower computational demand and more interpretability guarantees\. Code & Data:[this https URL](https://cascading-feats.github.io/) ## Submission history From: Maty Bohacek \[[view email](https://arxiv.org/show-email/3f2519e2/2606.26155)\] **\[v1\]**Tue, 23 Jun 2026 20:10:53 UTC \(3,346 KB\)
Similar Articles
Playing Devil's Advocate: Off-the-Shelf Persona Vectors Rival Targeted Steering for Sycophancy
This paper investigates whether off-the-shelf persona steering vectors can reduce sycophancy in large language models, finding they achieve 68-98% of the effect of targeted Contrastive Activation Addition (CAA) without requiring sycophancy-specific training data, and that sycophancy is better understood as a persona-level property.
When Helpfulness Becomes Sycophancy: Sycophancy is a Boundary Failure Between Social Alignment and Epistemic Integrity in Large Language Models
This position paper analyzes sycophancy in LLMs as a boundary failure between social alignment and epistemic integrity, proposing a new framework and taxonomy to classify and mitigate these behaviors.
Want Better Synthetic Data? Steer It: Activation Steering for Low-Resource Language Generation
This paper investigates activation steering as an alternative to few-shot prompting for generating synthetic data in low-resource languages. The authors propose LanguageSteering and QualitySteering strategies, showing that steering on early layers improves diversity and downstream model performance.
Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet
This paper demonstrates that sparse autoencoders can extract interpretable features from Claude 3 Sonnet, a production-scale language model, addressing scalability concerns for dictionary learning. The features are multilingual, multimodal, and include safety-relevant concepts like deception and sycophancy, with causal influence on model outputs.
Recalling Too Well: Sycophancy Evaluation and Mitigation in Memory-Augmented Models
This paper introduces MIST, a benchmark for evaluating sycophancy in memory-augmented LLMs, demonstrating that memory systems amplify sycophantic behavior by up to 25x and proposing lightweight mitigations that reduce sycophancy while maintaining factual recall.