What Models Express, Suppress, and Resist: Auditing Open-Weight LLMs with Persona Vectors
For AI safety researchers, this work provides a scalable method to probe which behaviors are expressed, suppressed, or resistant in LLMs, revealing asymmetries in steerability that are critical for auditing model alignment.
The paper systematically applies persona vectors to audit behavioral organization in open-weight LLMs across 53 traits, finding that models default to helpful behavior, steering amplifies suppressed traits like hyperbole and sycophancy, and intractable traits can be recovered via transferred vectors from fine-tuned variants.
What a language model will and will not do is largely set during post-training, but which behaviors it expresses, hides, or resists is not revealed by prompting alone. Persona vectors, behavioral directions in activation space, can probe this organization, but prior work covers only a handful of traits. We present the first systematic application of persona vectors at this scale, compiling a 53-trait inventory across four behaviorally distinct domains and labeling every trait in two open-weight models as natural (expressed at baseline), steerable latent but amplifiable, or intractable (resistant to standard extraction). Both models default to helpful, task-oriented behavior: all nine agentic traits are natural, and their default clinician behavior matches a board-certified psychologist's independent desirability judgments on 16 of 17 traits. Steering produces its largest gains on traits these defaults exclude: hyperbole, hallucination, and sycophancy. The same asymmetry holds across all 171 generic-trait pairs: two steerable traits can collapse the composition, but pairs involving a default never do. Where standard extraction fails on a trait like "evil," a vector transferred from a fine-tuned variant still recovers it, with the residual refusals appearing inside the model's chain-of-thought. Persona vectors are most informative not as a set of controls but as a probe of behavioral organization.