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 2511.04875cs.CLcs.AIcs.LG
收藏

LLM中行为自我意识的最小与机制性条件

Minimal and Mechanistic Conditions for Behavioral Self-Awareness in LLMs

University of California, San Diego(加州大学圣地亚哥分校);University of Virginia(弗吉尼亚大学);Durham University(杜ham大学)

Matthew Bozoukov, Matthew Nguyen, Shubkarman Singh, Bart Bussmann, Patrick Leask

AI总结:

本文通过LoRA微调实验发现LLM的行为自我意识可由单个秩-1适配器诱导,并能被激活空间中的单一引导向量捕获,且该能力具有领域局部性与线性特征。

摘要:

近期研究表明,大语言模型(LLMs)能够表现出行为自我意识:即在没有显式监督的情况下,准确描述或预测自身已学习行为的能力。这种能力引发了安全方面的担忧,例如它可能使模型在评估期间更好地隐藏其真实能力。我们试图刻画这种自我意识出现的最小条件,以及它得以表现的机制性过程。通过对指令微调的大语言模型进行低秩适配器(LoRA)的受控微调实验,我们发现:(1)使用单个秩-1 LoRA适配器即可可靠地诱导出自我意识;(2)所学习到的自我意识行为可以在激活空间中很大程度上被单个引导向量(steering vector)捕获,几乎恢复了微调带来的全部行为效应;(3)自我意识并非普遍存在,而是局限于特定领域,不同任务之间具有相互独立的表征。综合来看,这些发现表明行为自我意识是作为一种领域特定的线性特征出现的,并且可以很容易地被诱导和调节。

英文摘要:

Recent studies have revealed that LLMs can exhibit behavioral self-awareness: the ability to accurately describe or predict their own learned behaviors without explicit supervision. This capability raises safety concerns as it may, for example, allow models to better conceal their true abilities during evaluation. We attempt to characterize the minimal conditions under which such self-awareness emerges, and the mechanistic processes through which it manifests. Through controlled finetuning experiments on instruction-tuned LLMs with low-rank adapters (LoRA), we find: (1) that self-awareness can be reliably induced using a single rank-1 LoRA adapter; (2) that the learned self-aware behavior can be largely captured by a single steering vector in activation space, recovering nearly all of the fine-tune's behavioral effect; and (3) that self-awareness is non-universal and domain-localized, with independent representations across tasks. Together, these findings suggest that behavioral self-awareness emerges as a domain-specific, linear feature that can be easily induced and modulated.

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。

↑