cs.AINov 13, 2026

Safety Side Effects of Preventative Persona Steering: An Empirical Audit

Authors: Sabrina Tang, Nikolai Vo, Carson Murillo, Chira Sukumarnpiti, Sarah Wilson

Organizations: University of California, Irvine · University of Toledo · University of California, San Diego · Berkeley City College · 

Abstract

High-level behaviors such as sycophancy and hallucination correspond to directions in a model's activation space (Chen et al., 2025). These directions, known as persona vectors, can suppress undesirable traits either at inference time, subtracting the vector during generation, or through preventative steering, adding the vector during fine-tuning, so the optimizer never needs to encode the trait in weights, then removing the vector at evaluation. Chen et al. (2025) recommend preventative steering because it preserves Massive Multitask Language Understanding (MMLU) accuracy, but validations of this recommendation have measured capability alone. We primarily evaluate results on preventative steering of sycophancy and hallucination vectors on three models, measuring three safety benchmarks alongside trait expression, coherence, and MMLU. Our findings show that preventative steering still changes refusal behavior, but by an amount that depends on the trait and does not establish that preventative persona steering is safety-neutral, and steering methods should be evaluated on several safety benchmarks.

Figures & tables

Explore similar work

CardsList
  1. Persona Non Grata: Single-Method Safety Evaluation Is Incomplete for Persona-Imbued LLMs

    Apr 13, 2026Wenkai Li, Fan Yang, Shaunak A. Mehta +1Language Model Safety EvaluationLLM Safety

  2. Tracing Persona Vectors Through LLM Pretraining

    May 13, 2026Viktor Moskvoretskii, Dominik Glandorf, Jorge Medina Moreira +2Language Model PretrainingLLM Interpretability

  3. Persona Cartography: Charting Language Model Personality Traits in Weight Space

    Jul 8, 2026Luke Baines, Anton Gonzalvez Hawthorne, Mariia Koroliuk +4Personality Modeling in Language Models