arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2503.04429cs.AI
收藏

激活空间干预可在大语言模型之间迁移

Activation Space Interventions Can Be Transferred Between Large Language Models

Nvidia;Singapore University of Technology and Design(新加坡技术与设计大学)

Narmeen Oozeer, Dhruv Nathawani, Nirmalendu Prakash, Michael Lan, Abir Harrasse, Amirali Abdullah

AI总结:

本文提出通过激活空间映射在不同大语言模型间迁移安全干预,涵盖后门移除、有害提示拒绝和受损能力任务,并证明小模型可高效对齐大模型,自编码器映射可作为轻量级安全开关。

摘要:

对人工智能模型中表征普遍性的研究揭示了跨领域、跨模态和跨架构的日益融合。然而,表征普遍性的实际应用在很大程度上仍未得到探索。我们通过证明安全干预可以通过学习共享激活空间的映射在模型之间迁移,来弥合这一差距。我们在两个成熟的人工智能安全任务上展示了这一方法:后门移除和拒绝有害提示,结果表明能够成功迁移以可预测方式改变模型输出的引导向量。此外,我们提出了一个新任务,即“受损能力”,其中模型经过微调以嵌入与后门相关的知识。这测试了它们将有用技能与后门分离的能力,反映了现实世界的挑战。在 Llama、Qwen 和 Gemma 模型系列上的大量实验表明,我们的方法能够使用较小的模型高效地对齐较大的模型。此外,我们证明基础模型与微调模型之间的自编码器映射可以作为可靠的“轻量级安全开关”,允许在模型行为之间动态切换。

英文摘要:

The study of representation universality in AI models reveals growing convergence across domains, modalities, and architectures. However, the practical applications of representation universality remain largely unexplored. We bridge this gap by demonstrating that safety interventions can be transferred between models through learned mappings of their shared activation spaces. We demonstrate this approach on two well-established AI safety tasks: backdoor removal and refusal of harmful prompts, showing successful transfer of steering vectors that alter the models' outputs in a predictable way. Additionally, we propose a new task, \textit{corrupted capabilities}, where models are fine-tuned to embed knowledge tied to a backdoor. This tests their ability to separate useful skills from backdoors, reflecting real-world challenges. Extensive experiments across Llama, Qwen and Gemma model families show that our method enables using smaller models to efficiently align larger ones. Furthermore, we demonstrate that autoencoder mappings between base and fine-tuned models can serve as reliable ``lightweight safety switches", allowing dynamic toggling between model behaviors.

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。

↑