Decomposing Refusal Steering in Mixture-of-Experts Models
TL;DR
This work extends this refusal steering method to three open-source Mixture-of-Experts (MoE) LLMs to decompose and better understand how refusal mechanisms operate across MoE components, finding that steering performance is uninhibited by the complex routing patterns inherent to the MoE architecture.
Abstract
Safety alignment in instruction-tuned large language models (LLMs) depends on a model's ability to reliably refuse harmful or disallowed requests. Recent work has shown that a steering vector can be applied to a dense LLM during inference to suppress refusal behavior and induce responses to harmful requests. We extend this refusal steering method to three open-source Mixture-of-Experts (MoE) LLMs to decompose and better understand how refusal mechanisms operate across MoE components. We find that steering performance is uninhibited by the complex routing patterns inherent to the MoE architecture, and that a single expert recovers 78% of the full steering effect on average when free to select its own location. However, expert-level steering only recovers 54% on average when constrained to the location used by full-layer steering, a gap that widens further under safety-related system prompts. Our results also show that refusal signals captured by these steering methods differ from those captured by expert routing behavior. Together, these results suggest that refusal behavior is distributed unevenly across the MoE LLM architecture rather than governed by a single unified mechanism.
