让语言模型更精准地抑制不良行为，只干预需要调整的神经元。
GAPS: Dimension-Level Gates for Conditional Activation Steering
- 按神经元维度判断是否干预，比全局干预更精细。
- 在毒性去除任务中，将模型毒性率从6.52%降到0.48%。
- 无需训练，可无缝接入现有调控方法，适合安全增强场景。
激活调控通过生成时向隐藏状态添加调控向量来抑制语言模型的不良行为。近期条件化方法如CAST和DSAS通过决定何时干预提升了行为与能力的权衡，但一旦启动，仍对所有隐藏维度施加完整的密集向量，无论该神经元是否携带概念信息或已处于期望状态。本文提出维度级条件化作为互补的细粒度选择机制，引入GAPS（基于后验与可分性的门控激活调控）：两个无需训练的门控机制——静态可分性门通过AUROC筛选具备统计可靠概念信息的神经元，动态后验门仅在当前激活更符合不良概念（高斯模型下）时才进行干预。每令牌开销约O(D)，可集成至现有方法。在Gemma-3（4B）与Qwen-3（1.7B）上，针对RealToxicityPrompts（毒性缓解）与OneSeC（概念移除）任务，GAPS始终达到或超越其令牌级对应方法的帕累托前沿；固定能力预算下，DSAS+GAPS将Gemma-3的毒性率由6.52%降至0.48%，优于DSAS单独使用的3.52%。消融实验表明主要增益来自后验门。
原文摘要 · Abstract (English)
Activation steering suppresses undesired behaviors in language models by adding a steering vector to the hidden state during generation. Recent conditional methods such as CAST and DSAS improve the behavior-capability trade-off by deciding when to intervene, but once active, they apply the full dense vector to all hidden dimensions, regardless of whether a neuron carries concept information or already lies in the desired regime. We introduce dimension-level conditioning as a complementary axis of selectivity that also decides which neurons to intervene on. Our method, GAPS (Gated Activation steering via Posterior and Separability), combines two training-free gates: a static separability gate that restricts steering to neurons with statistically reliable concept information (via AUROC), and a dynamic posterior gate that steers a neuron only when its current activation is better explained by the undesired concept under a Gaussian model. The gates add O(D) overhead per token, and they plug into existing conditional methods. On toxicity mitigation (RealToxicityPrompts) and concept removal (OneSeC) with Gemma-3 (4B) and Qwen-3 (1.7B), GAPS consistently matches or improves the Pareto front of its token-level counterparts; under a fixed capability budget, DSAS+GAPS reduces Gemma-3's toxicity rate from 6.52% to 0.48%, versus 3.52% for DSAS alone. Ablations attribute most of the gain to the posterior gate.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。