arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2605.29971cs.CL
收藏

连续变量的因果干预:以上下文学习中转向向量的动词偏向为例

Causal Interventions on Continuous Variables: A Case Study on Verb Bias in Steering Vectors for In-Context Learning

Yale University(耶鲁大学)

Zhenghao Herbert Zhou, R. Thomas McCoy, Robert Frank

AI总结:

提出一种对连续变量进行因果干预的方法,通过定位低维方向并编辑向量实现反事实目标值,应用于动词偏向特征,证明其在语言模型中的因果表示,并探讨与上下文学习的关系。

摘要:

语言模型表示中的因果干预主要针对离散特征,如语法数。然而,语言模型也必须利用分级特征。我们引入了一种对连续变量进行因果干预的方法:给定与分级目标变量配对的激活向量,我们定位该变量的低维方向,并使用该方向将向量编辑为反事实目标值。我们将此方法应用于心理语言学中研究充分的连续特征,即动词偏向(反映给定动词后倾向于出现哪种句法结构)。我们表明,动词偏向因果地表示在从大型语言模型中提取的转向向量中:对动词偏向的反事实编辑系统地改变了下游结构偏好。动词偏向此前也与上下文学习相关联;在进一步分析中,我们发现转向向量编码了可能驱动上下文学习中观察到的误差驱动更新行为的误差信号,但这些转向向量的方面在下游生成中并未被因果使用。总体而言,这些结果表明因果干预可以应用于连续变量,尽管将连续变量与上下文学习联系起来仍然是一个挑战。

英文摘要:

Causal interventions in language model representations have largely targeted discrete features, like grammatical number. However, language models must also make use of features that are graded. We introduce a method for causal intervention on continuous variables: given activation vectors paired with a graded target variable, we localize a low-dimensional direction for that variable and use this direction to edit a vectors toward counterfactual target values. We apply this method to a continuous feature that is well-studied in psycholinguistics, namely verb bias (which reflects which syntactic structures tend to follow a given verb). We show that verb bias is causally represented in steering vectors extracted from large language models: counterfactual edits to verb bias systematically shift downstream structural preferences. Verb bias has also previously been linked to in-context learning; in further analyses, we find that steering vectors encode error signals that could drive the error-driven update behavior seen in in-context learning but that these aspects of the steering vectors are not causally used in downstream production. Overall, these results show causal interventions can be applied to continuous variables, though connecting continuous variables to in-context learning remains a challenge.

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。

↑