Search papers, labs, and topics across Lattice.
This paper introduces a method for causally intervening on continuous variables in LLM activations by localizing a low-dimensional direction corresponding to a graded target variable and editing vectors along this direction. They apply this method to verb bias, a continuous feature reflecting syntactic structure preferences, and demonstrate that counterfactual edits to verb bias in steering vectors systematically shift downstream structural preferences. Further analysis reveals that while steering vectors encode error signals related to in-context learning, these signals are not causally utilized in downstream production.
Editing continuous features like "verb bias" in LLM steering vectors can predictably shift downstream syntactic preferences, but the link to in-context learning remains elusive.
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.
