@inproceedings{huang-etal-2026-role,
    title = "Role-Sensitive Neurons: A Neuron-Level Gain Control Mechanism for Confidence Steering",
    author = "Huang, Peiwen  and
      Hsu, Chih-Hao  and
      Huang, Tzu-Hung  and
      Lin, Shou-De",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.findings-acl.294/",
    pages = "5924--5944",
    ISBN = "979-8-89176-395-1",
    abstract = "Role-playing prompts effectively steer Large Language Models (LLMs), yet the neural mechanism driving this behavioral shift remains unclear. In this work, we identify Role-Sensitive Neurons (RSNs){---}a sparse sub-network ({\ensuremath{\approx}} 0.5{\%} of all neurons) governing the transition from hesitation to action. Using a novel evaluation framework with explicit abstention (MMLU-E), we reveal a Confidence-Performance Decoupling: roles primarily modulate the model{'}s probabilistic ``willingness to act'' rather than its underlying knowledge representation. We demonstrate that RSNs function as a mechanistic gain control system: causal intervention on this subspace allows precise regulation of abstention behavior. Furthermore, cross-model transfer experiments confirm that these circuits are indigenous to pre-training, with Instruction Tuning (SFT) acting merely as a ``signal sharpener'' to refine latent gain dynamics. Finally, we identify a critical safety boundary: in knowledge-deficient models, amplifying RSNs induces ``unwarranted certainty,'' highlighting decisiveness as a tunable gain parameter distinct from epistemic truth."
}