Cognitive Behavior Modeling via Activation Steering

Anthony Kuang, Ahmed Ismail, Ayo Akinkugbe, Kevin Zhu, Sean O'Brien

Keywords: Activation Steering, Steering Vectors, Mechanistic Interpretability

TL;DR: CBMAS is a diagnostic framework for fine-grained interpretation of cognitive behaviors in LLMs by continuously steering activations and analyzing how intervention strength and layer depth affect model biases.

Abstract: Large language models (LLMs) often encode cognitive behaviors unpredictably across prompts, layers, and contexts, making them difficult to diagnose and control. We present $\textbf{CBMAS}$, a diagnostic framework for continuous activation steering, which extends cognitive bias analysis from discrete before/after interventions to interpretable trajectories. By combining steering vector construction with dense $\alpha$-sweeps, logit lens-based bias curves, and layer-site sensitivity analysis, our approach can reveal tipping points where small intervention strengths flip model behavior and show how steering effects evolve across layer depth. We argue that these continuous diagnostics offer a bridge between high-level behavioral evaluation and low-level representational dynamics, contributing to the cognitive interpretability of LLMs. Lastly, we provide a CLI and datasets for various cognitive behaviors at the project repository, https://github.com/shimamooo/CBMAS.

Submission Number: 118