Steering Llama-2-7B-Chat and Qwen2.5-7B-Instruct teachers and distilling students on benign data transfers measurable jailbreak susceptibility, with Llama showing threshold behavior at α = -0.15 and Qwen reaching transfer ratios up to 0.61.
Playing Devil's Advocate: Off-the-Shelf Persona Vectors Rival Targeted Steering for Sycophancy
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
We study the effect of different persona on \textbf{sycophancy}: model's agreement with users even when the user is incorrect. The standard mitigation, Contrastive Activation Addition (CAA), derives a steering direction from labelled pairs of sycophantic and honest responses. This study evaluates whether off-the-shelf persona steering vectors, originally developed for general role-playing and not trained on sycophancy data, can serve as an alternative. In two instruction-tuned models, steering toward personas characterised by doubt or scrutiny reduces sycophancy to approximately $68\%$ and $98\%$ of CAA's effect, and, unlike CAA, maintains accuracy when the user is correct. The effect is also asymmetric: steering toward agreeable personas does not produce a mirror increase in sycophancy. Geometrically, the persona vector is largely independent of the direction of sycophancy in activation space. Collectively, these findings suggest that sycophancy is better understood as a persona-level property rather than a single steerable direction. We release our code here: https://anonymous.4open.science/r/Sycophancy-Steering-9DF0/.
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Quantifying Subliminal Behavioral Transfer Ratios in Language Model Distillation
Steering Llama-2-7B-Chat and Qwen2.5-7B-Instruct teachers and distilling students on benign data transfers measurable jailbreak susceptibility, with Llama showing threshold behavior at α = -0.15 and Qwen reaching transfer ratios up to 0.61.