{"ID":3053213,"CreatedAt":"2026-06-04T04:41:36.695875263Z","UpdatedAt":"2026-06-05T19:19:17.853951865Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2606.04160","arxiv_id":"2606.04160","title":"Expert-Aware Refusal Steering","abstract":"Safety alignment in instruction-tuned large language models (LLMs) depends on a model's ability to reliably refuse to respond to harmful or disallowed requests. Recent work has shown that a steering vector can be applied to a dense LLM during inference to effectively suppress refusal behavior, inducing response to harmful requests. We extend this refusal steering method to three open-source Mixture-of-Experts (MoE) LLMs and find that steering performance is uninhibited by the complex routing patterns inherent to the MoE architecture. We then propose two expert-aware refusal steering methods that leverage refusal-specific expert routing patterns and expert-specific steering directions to suppress normal refusal behavior. We find that refusal behavior can be effectively steered based on the output of a single expert. Our results show that refusal signals captured by steering methods differ from expert routing behavior, suggesting a substantial role for attention in MoE refusal behavior.","short_abstract":"Safety alignment in instruction-tuned large language models (LLMs) depends on a model's ability to reliably refuse to respond to harmful or disallowed requests. Recent work has shown that a steering vector can be applied to a dense LLM during inference to effectively suppress refusal behavior, inducing response to harm...","url_abs":"https://arxiv.org/abs/2606.04160","url_pdf":"https://arxiv.org/pdf/2606.04160v1","authors":"[\"Anna C. Marbut\",\"Daniel R. Olson\",\"Travis J. Wheeler\"]","published":"2026-06-02T19:23:46Z","proceeding":"cs.CL","tasks":"[\"cs.CL\",\"cs.LG\"]","methods":"[\"Large Language Model\",\"Language Model\"]","has_code":false}