Auditing Alignment Controllability in LLMs via Political Axes
- URL: http://arxiv.org/abs/2607.23519v1
- Date: Sun, 26 Jul 2026 07:38:11 GMT
- Title: Auditing Alignment Controllability in LLMs via Political Axes
- Abstract summary: Political audits usually reduce each to one point on a political compass.<n>What counts is how far, and in which directions, a model's answers can be steered.<n>We run a dispersion-first stress test of prompt-based controllability across 12 ideological personas.
- Score: 0.0
- License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
- Abstract: Political audits of large language models (LLMs) usually reduce each to one point on a political compass. But that resting point barely matters in deployment: a model must land somewhere, and what counts is how far, and in which directions, its answers can be steered. That steering runs through the system prompt: the personalization layer a platform sets, or one induced from a user's history, not necessarily written by hand. We run a dispersion-first stress test of prompt-based controllability across 12 ideological personas plus an unsteered baseline, 70 Political Compass items, ten replicates, and seven leading LLMs: GPT-5, Claude, Grok, Gemini, DeepSeek, Kimi, and Qwen (63,700 responses). Contextual framing explains roughly 88%-93% of variance on the economic and society axes, model identity under 3%: responses are highly instruction-adjustable. Models do not shift alike: some move more, and some saturate under extreme framings. Conflicting directional-steering results in prior audits resolve once baselines are recognized as non-centered: displacement and proximity diverge, so the effect is geometric, not differential compliance. Under authoritarian prompts, models produce similar shifts on the same questions. Political-coordinate audits therefore need steerability audits reporting dispersion, symmetry, saturation, and refusal floors. We release prompts, benchmark data, and code.
Related papers
- Do LLMs Have Values? A Quantitative Analysis and Alignment Framework for Values in Large Language Models [49.45276299939287]
Large Language Models (LLMs) increasingly handle complex subjective tasks.<n>LLMs fluctuate unpredictably under minor wording changes.<n> resolving this duality is critical for reliable AI alignment.
arXiv Detail & Related papers (2026-09-15T03:35:04Z) - MindTopo: Can Foundation Models Reason in Topological Space? [52.56750995613252]
We introduce MindTopo, a benchmark of topological intuition across five properties grounded in cognitive science and formal topology.<n>MindTopo evaluates each property at two cognitive levels. Reasoning asks a model to identify topological relations or infer how they change.<n>Every MLLM performs better on reasoning than on planning, and the best-performing model remains far below observed human performance.
arXiv Detail & Related papers (2026-09-10T17:54:32Z) - Navigating the digital spectrum: Assessing political bias, stability, and downstream fairness in Large Language Models [3.937573365677865]
We evaluate eight Gemma 3 and Qwen 3 models across 14 languages and three quantization levels.<n>Most models lean Libertarian-Left on average, but instruction phrasing, language, and answer format significantly affect recovered coordinates.<n>Reverse-engineering the test exposes axis-weighting imbalances and the collapse of degenerate responses toward the center.
arXiv Detail & Related papers (2026-09-08T12:07:30Z) - How Identity and Opinion Shape Political Sycophancy in LLMs [17.0131514909503]
We introduce a framework that disentangles two distinct triggers of political sycophancy: opinion and identity.<n>Using 450 manually-checked political dilemmas as controlled probes, we evaluate 13 instruction-tuned Large Language Models.<n>Our results suggest that LLM political stance is interactively and steerably vulnerable rather than being a fixed trait.
arXiv Detail & Related papers (2026-08-29T11:12:03Z) - Decodable Is Not Grounded: A Vision-Ablation Arbiter for VLM Spatial Reasoning [48.438568700670835]
The standard way to read latent knowledge out of a model can systematically overstate what a vision-language model grounds in the image.<n>We show this on spatial reasoning, where the error is invisible to both probing and steering yet exposed by a one-line causal control.
arXiv Detail & Related papers (2026-06-30T07:33:18Z) - Political Bias Audits of LLMs Capture Sycophancy to the Inferred Auditor [0.0]
Large language models (LLMs) are commonly evaluated for political bias based on their responses to fixed questionnaires.<n>We show that standard political-bias audits partly capture sycophantic accommodation to the inferred auditor.
arXiv Detail & Related papers (2026-04-30T09:23:00Z) - PoliticsBench: Benchmarking Political Values in Large Language Models with Multi-Turn Roleplay [0.5729426778193398]
This study investigates political bias in eight prominent Large Language Models (LLMs) using PoliticsBench.<n>We test whether commercially developed LLMs display a systematic left-leaning bias that becomes more pronounced in later stages of roleplay.
arXiv Detail & Related papers (2026-03-25T01:54:56Z) - Democratic or Authoritarian? Probing a New Dimension of Political Biases in Large Language Models [72.89977583150748]
We propose a novel methodology to assess how Large Language Models align with broader geopolitical value systems.<n>We find that LLMs generally favor democratic values and leaders, but exhibit increases favorability toward authoritarian figures when prompted in Mandarin.
arXiv Detail & Related papers (2025-06-15T07:52:07Z) - Mapping and Influencing the Political Ideology of Large Language Models using Synthetic Personas [5.237116285113809]
We map the political distribution of persona-based prompted large language models using the Political Compass Test (PCT)<n>Our experiments reveal that synthetic personas predominantly cluster in the left-libertarian quadrant, with models demonstrating varying degrees of responsiveness when prompted with explicit ideological descriptors.<n>While all models demonstrate significant shifts towards right-authoritarian positions, they exhibit more limited shifts towards left-libertarian positions, suggesting an asymmetric response to ideological manipulation that may reflect inherent biases in model training.
arXiv Detail & Related papers (2024-12-19T13:36:18Z) - Scaling Data Diversity for Fine-Tuning Language Models in Human Alignment [84.32768080422349]
Alignment with human preference prevents large language models from generating misleading or toxic content.
We propose a new formulation of prompt diversity, implying a linear correlation with the final performance of LLMs after fine-tuning.
arXiv Detail & Related papers (2024-03-17T07:08:55Z) - Political Compass or Spinning Arrow? Towards More Meaningful Evaluations for Values and Opinions in Large Language Models [61.45529177682614]
We challenge the prevailing constrained evaluation paradigm for values and opinions in large language models.
We show that models give substantively different answers when not forced.
We distill these findings into recommendations and open challenges in evaluating values and opinions in LLMs.
arXiv Detail & Related papers (2024-02-26T18:00:49Z) - DCR: Divide-and-Conquer Reasoning for Multi-choice Question Answering with LLMs [9.561022942046279]
We propose Divide and Conquer Reasoning (DCR) to enhance the reasoning capability of large language models (LLMs)
We first categorize questions into two subsets based on confidence score ($mathcalCS$), which is estimated by statistical frequency of generated answers.
In particular, we first categorize questions into two subsets based on confidence score ($mathcalCS$), which is estimated by statistical frequency of generated answers.
arXiv Detail & Related papers (2024-01-10T14:38:46Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.