When Models Refuse: Political Steerability and Feature Richness as Measures of Ideological Depth
On the map: LLM safety and security → Safety alignment, refusal and guardrails
also in 1 more
Abstract (arXiv)
Large language models (LLMs) sometimes refuse to follow benign instructions, such as declining to argue a political position or adopt a stated persona, and such refusals are commonly read as safety guardrails at work. We ask whether they can instead signal a **capability deficit**: a shortage of the internal representations a model needs to reason from the instructed perspective. To investigate, we introduce **ideological depth**, a property with two components: (i) a model's ability to follow political instructions without *failure* (steerability), and (ii) the **feature richness** of its internal political representations, measured with sparse autoencoders (SAEs). Using two widely used openweight LLMs as candidates, we compare interventions based on prompts and activation-steering, and probe political features with publicly available SAEs. We find large, systematic differences: a model that is more steerable in both ideological directions activates **~7.3x** more distinct political features, while the other model instead responds with increased refusals. Causally ablating a small, targeted set of political features from the former model reproduces the same feature-poor behavior and drives up refusals. Together, these results indicate that refusals on benign prompts can arise from **capability deficits** rather than fixed safety rules, and that ideological depth is a measurable property of LLMs that helps predict when a model will refuse.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Safety alignment, refusal and guardrails, one of 12 directions in LLM safety and security. Builds on 13 of its 28 references that are in the map. First posted on arXiv 29 Aug 2025.
Builds on
13 of 28 references.
References · 13
- 2016Understanding intermediate layers using linear classifier probescited by 1,353
- 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,085
- 2016Semantics derived automatically from language corpora contain human-like biasescited by 901
- 2022Discovering Language Model Behaviors with Model-Written Evaluationscited by 836
- 2023Whose Opinions Do Language Models Reflect?cited by 773
- 2023Large Language Models Are Not Robust Multiple Choice Selectorscited by 486
- 2024Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2cited by 394
- 2025Persona Vectors: Monitoring and Controlling Character Traits in Language Modelscited by 345
- 2024Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Values and Opinions in Large Language Modelscited by 162
- 2024Llama Scope: Extracting Millions of Features from Llama-3.1-8B with Sparse Autoencoderscited by 143
- 2025SAEs Are Good for Steering -- If You Select the Right Featurescited by 78
- 2025Beyond Prompt Engineering: Robust Behavior Control in LLMs via Steering Target Atomscited by 24
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026Political Alignment in Large Language Models: A Multidimensional Audit of Psychometric Identity and Behavioral Bias
- 2026Navigating the digital spectrum: Assessing political bias, stability, and downstream fairness in Large Language Models
- 2025Steering Towards Fairness: Mitigating Political Bias in LLMs
- 2025Examining Alignment of Large Language Models through Representative Heuristics: The Case of Political Stereotypes
- 2026LLM-Ideoplasticity: Measuring Ideological Plasticity in the Political Behavior of LLMs as a Context-Conditioned Distribution
- 2024GermanPartiesQA: Benchmarking Commercial Large Language Models and AI Companions for Political Alignment and Sycophancy
- 2026Auditing Alignment Controllability in LLMs via Political Axes
- 2025Linear Representations of Political Perspective Emerge in Large Language Models