语言模型引导的圆柱表示假说
The Cylindrical Representation Hypothesis for Language Model Steering
AI总结:
提出圆柱表示假说(CRH),通过放宽线性表示假说(LRH)的正交性假设,解释语言模型引导中的不稳定性和不确定性。
摘要:
引导是一种广泛用于控制大型语言模型的技术,但其效果往往不稳定且难以预测。现有的理论解释主要基于线性表示假说(LRH)。虽然LRH假设概念可以正交化以实现无损控制,但这种理想化的映射在真实表示中无法实现,也无法解释观察到的引导不可预测性。通过放宽LRH的正交性假设同时保留线性表示,我们展示了重叠的概念贡献自然产生一种样本特定的轴正交结构。我们将此形式化为圆柱表示假说(CRH)。在CRH中,中心轴捕捉概念缺失与存在之间的主要差异,并驱动概念生成。周围的法平面通过决定轴激活目标概念的难易程度来控制引导敏感性。在该平面内,只有特定的敏感扇区强烈促进概念激活,而其他扇区可能抑制或延迟激活。虽然周围的法平面可以从差异向量中可靠识别,但敏感扇区无法识别,从而在扇区层面引入内在不确定性。这种不确定性提供了原则性解释,说明为什么即使使用良好对齐的方向,引导结果也常常波动。我们的实验验证了圆柱结构的存在,并证明CRH为解释真实场景中的模型引导行为提供了一种有效且实用的方法:https://github.com/mbzuai-nlp/CRH。
英文摘要:
Steering is a widely used technique for controlling large language models, yet its effects are often unstable and hard to predict. Existing theoretical accounts are largely based on the Linear Representation Hypothesis (LRH). While LRH assumes that concepts can be orthogonalized for lossless control, this idealized mapping fails in real representations and cannot account for the observed unpredictability of steering. By relaxing LRH's orthogonality assumption while preserving linear representations, we show that overlapping concept contributions naturally yield a sample-specific axis-orthogonal structure. We formalize this as the Cylindrical Representation Hypothesis (CRH). In CRH, a central axis captures the main difference between concept absence and presence and drives concept generation. A surrounding normal plane controls steering sensitivity by determining how easily the axis can activate the target concept. Within this plane, only specific sensitive sectors strongly facilitate concept activation, while other sectors can suppress or delay it. While the surrounding normal plane can be reliably identified from difference vectors, the sensitive sector cannot, introducing intrinsic uncertainty at the sector level. This uncertainty provides a principled explanation for why steering outcomes often fluctuate even when using well-aligned directions. Our experiments verify the existence of the cylindrical structure and demonstrate that CRH provides a valid and practical way to interpret model steering behavior in real settings: https://github.com/mbzuai-nlp/CRH.
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。