Sycophancy Hides Linearly in the Attention Heads

TL;DR

This paper investigates how sycophantic behavior in language models is linearly separable within attention heads, revealing that targeted linear interventions can mitigate such biases by exploiting internal attention patterns.

Contribution

It demonstrates that sycophancy signals are most effectively manipulated within specific attention heads and introduces linear probing methods to analyze and reduce this behavior.

Findings

Sycophancy signals are linearly separable in attention heads.

Probes trained on TruthfulQA transfer to other factual benchmarks.

Attention heads attend disproportionately to expressions of user doubt.

Abstract

We find that correct-to-incorrect sycophancy signals are most linearly separable within multi-head attention activations. Motivated by the linear representation hypothesis, we train linear probes across the residual stream, multilayer perceptron (MLP), and attention layers to analyze where these signals emerge. Although separability appears in the residual stream and MLPs, steering using these probes is most effective in a sparse subset of middle-layer attention heads. Using TruthfulQA as the base dataset, we find that probes trained on it transfer effectively to other factual QA benchmarks. Furthermore, comparing our discovered direction to previously identified "truthful" directions reveals limited overlap, suggesting that factual accuracy, and deference resistance, arise from related but distinct mechanisms. Attention-pattern analysis further indicates that the influential heads attend disproportionately to expressions of user doubt, contributing to sycophantic shifts. Overall, these findings suggest that sycophancy can be mitigated through simple, targeted linear interventions that exploit the internal geometry of attention activations.