Dissecting Persona-Driven Reasoning in Language Models via Activation Patching
This work addresses the problem of understanding persona-driven reasoning in LLMs for researchers and practitioners, but it is incremental as it takes a first step using activation patching without achieving broad SOTA or proof.
The study investigated how assigning personas influences language models' reasoning on objective tasks, finding that early MLP layers process both syntax and semantics to transform persona tokens, while middle MHA layers use these to shape outputs, with specific attention heads disproportionately attending to racial and color-based identities.
Large language models (LLMs) exhibit remarkable versatility in adopting diverse personas. In this study, we examine how assigning a persona influences a model's reasoning on an objective task. Using activation patching, we take a first step toward understanding how key components of the model encode persona-specific information. Our findings reveal that the early Multi-Layer Perceptron (MLP) layers attend not only to the syntactic structure of the input but also process its semantic content. These layers transform persona tokens into richer representations, which are then used by the middle Multi-Head Attention (MHA) layers to shape the model's output. Additionally, we identify specific attention heads that disproportionately attend to racial and color-based identities.