# What Models Express, Suppress, and Resist: Auditing Open-Weight LLMs with Persona Vectors

arXiv: https://arxiv.org/abs/2607.13162 · HTML: https://mlsanity.com/paper/2607.13162/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 59 references that are in the map.
- First posted on arXiv 14 Jul 2026.

## Builds on (most cited first, 23 of 23)

- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [A Survey on Large Language Model based Autonomous Agents](https://mlsanity.com/paper/2308.11432/) (2023)
- [The Rise and Potential of Large Language Model Based Agents: A Survey](https://mlsanity.com/paper/2309.07864/) (2023)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [Steering Llama 2 via Contrastive Activation Addition](https://mlsanity.com/paper/2312.06681/) (2023)
- [Towards Understanding Sycophancy in Language Models](https://mlsanity.com/paper/2310.13548/) (2023)
- [Steering Language Models With Activation Engineering](https://mlsanity.com/paper/2308.10248/) (2023)
- [Discovering Language Model Behaviors with Model-Written Evaluations](https://mlsanity.com/paper/2212.09251/) (2022)
- [Persona Vectors: Monitoring and Controlling Character Traits in Language Models](https://mlsanity.com/paper/2507.21509/) (2025)
- [Character-LLM: A Trainable Agent for Role-Playing](https://mlsanity.com/paper/2310.10158/) (2023)
- [Deliberative Alignment: Reasoning Enables Safer Language Models](https://mlsanity.com/paper/2412.16339/) (2024)
- [Two Tales of Persona in LLMs: A Survey of Role-Playing and Personalization](https://mlsanity.com/paper/2406.01171/) (2024)
- [A Computational Approach to Understanding Empathy Expressed in Text-Based Mental Health Support](https://mlsanity.com/paper/2009.08441/) (2020)
- [The Assistant Axis: Situating and Stabilizing the Default Persona of Language Models](https://mlsanity.com/paper/2601.10387/) (2026)
- [Sycophancy Hides Linearly in the Attention Heads](https://mlsanity.com/paper/2601.16644/) (2026)
- [Personality Vector: Modulating Personality of Large Language Models by Model Merging](https://mlsanity.com/paper/2509.19727/) (2025)
- [BILLY: Steering Large Language Models via Merging Persona Vectors for Creative Generation](https://mlsanity.com/paper/2510.10157/) (2025)
- [Crafting Customisable Characters with LLMs: A Persona-Driven Role-Playing Agent Framework](https://mlsanity.com/paper/2406.17962/) (2024)
- [Steering at the Source: Style Modulation Heads for Robust Persona Control](https://mlsanity.com/paper/2603.13249/) (2026)
- [Evaluating the Clinical Safety of LLMs in Response to High-Risk Mental Health Disclosures](https://mlsanity.com/paper/2509.08839/) (2025)
- [Letting Tutor Personas Speak Up for LLMs: Learning Steering Vectors from Dialogue via Preference Optimization](https://mlsanity.com/paper/2602.07639/) (2026)
- [Bridging Mechanistic Interpretability and Prompt Engineering with Gradient Ascent for Interpretable Persona Control](https://mlsanity.com/paper/2601.02896/) (2026)
- [Persona Jailbreaking in Large Language Models](https://mlsanity.com/paper/2601.16466/) (2026)

## Abstract (arXiv)

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.