arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.12158cs.LG
收藏

SafeNeuron: 大语言模型的神经层面安全对齐

SafeNeuron: Neuron-Level Safety Alignment for Large Language Models

Xidian University, Xi'an, China(西安电子科技大学);Harbin Institute of Technology, Harbin, China(哈尔滨工业大学);National University of Singapore,Singapore(新加坡国立大学)

Zhaoxin Wang, Jiaming Liang, Fengbin Zhu, Weixiang Zhao, Junfeng Fang, Jiayi Ji, Handing Wang, Tat-Seng Chua

AI总结:

SafeNeuron通过神经层面安全对齐框架提升模型鲁棒性,减少攻击风险并保持通用能力。

摘要:

大语言模型(LLMs)和多模态LLMs在发布前通常会进行安全对齐以防止生成有害内容。然而,近期研究表明,安全行为集中在参数的小子集中,使对齐变得脆弱,容易通过神经层面攻击绕过。此外,大多数现有对齐方法在行为层面操作,对模型内部安全机制的控制有限。在本工作中,我们提出了SafeNeuron,一种神经层面的安全对齐框架,通过在网络中重新分布安全表示来提高鲁棒性。SafeNeuron首先识别与安全相关的神经元,然后在偏好优化期间冻结这些神经元,以防止依赖稀疏的安全路径,并迫使模型构建冗余的安全表示。在跨模型和多模态的广泛实验中,SafeNeuron显著提高了对神经剪枝攻击的鲁棒性,降低了开源模型被重新利用为红队生成器的风险,并保持了通用能力。此外,我们的逐层分析揭示安全行为由稳定且共享的内部表示所支配。总体而言,SafeNeuron为模型对齐提供了可解释且稳健的视角。

英文摘要:

Large language models (LLMs) and multimodal LLMs are typically safety-aligned before release to prevent harmful content generation. However, recent studies show that safety behaviors are concentrated in a small subset of parameters, making alignment brittle and easily bypassed through neuron-level attacks. Moreover, most existing alignment methods operate at the behavioral level, offering limited control over the model's internal safety mechanisms. In this work, we propose SafeNeuron, a neuron-level safety alignment framework that improves robustness by redistributing safety representations across the network. SafeNeuron first identifies safety-related neurons, then freezes these neurons during preference optimization to prevent reliance on sparse safety pathways and force the model to construct redundant safety representations. Extensive experiments across models and modalities demonstrate that SafeNeuron significantly improves robustness against neuron pruning attacks, reduces the risk of open-source models being repurposed as red-team generators, and preserves general capabilities. Furthermore, our layer-wise analysis reveals that safety behaviors are governed by stable and shared internal representations. Overall, SafeNeuron provides an interpretable and robust perspective for model alignment.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑