Incoherent Values? Probing LLM Preferences Through Parametric Variation

• URL: http://arxiv.org/abs/2606.21102v1
• Date: Fri, 19 Jun 2026 05:07:34 GMT
• Title: Incoherent Values? Probing LLM Preferences Through Parametric Variation
• Abstract summary: It is hard to trust someone with incoherent values.<n>Some think of Large Language Models as merely text generators with no evaluative core.<n>We show that models given time to reason are less incoherent than those with thinking disabled.
• Score: 0.1753733541634709
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: To trust another autonomous entity -- human or AI -- it helps to know that how it acts given one set of reasons is at least somewhat predictive of how it would act under others. It is hard to trust someone with incoherent values. Some think of Large Language Models as merely stochastic text generators with no evaluative core -- superpositions of billions of possible characters, not one stable identity. But others have argued that LLMs *do* have stable, emergent values, which can be elicited by presenting them with a series of forced choices between arbitrary statements, and which emerge as a function of model scale. In this paper, we test this thesis by presenting LLMs with parametric variations on those forced choices. We reason that if a model genuinely prefers A to B, then except in unusual circumstances it should also reject B in favor of an augmented version of A, which has more of what makes A desirable -- A++. Our results indicate that earlier attributions of coherence may have overstated their case. Even the most capable models exhibit significant incoherence, and coherence does not appear to emerge as a result of underlying model capability. We do, however, find that models given time to reason are less incoherent than those with thinking disabled. More generally, we develop a novel framework for eliciting and evaluating coherent values, which can be used both to assess how trustworthy current models are, and -- in future work -- to provide reward signal that can be used for making more coherent agents.

Related papers

• An Empirical Study of Counterfactual Self-Explanations in LLMs [13.093234754814851]
  Large language models can easily generate explanations for their own outputs, but such self-explanations are not necessarily faithful to the model's behavior.<n>We study this issue through counterfactual self-explanations, where a model minimally edits an input so that its own prediction changes.<n>Our results show that model scale is the strongest determinant of explanation quality.
  arXiv  Detail & Related papers  (2026-09-15T12:47:22Z)
• Position: It's Time to Optimize LLMs for Self-Consistency [50.68587056599734]
  Many important failures persist in language model pre- and post-training pipelines.<n>We argue that these failures arise from a modeling assumption permeating all aspects of the pipeline.<n>We propose self-consistency as a framework for understanding these failures.
  arXiv  Detail & Related papers  (2026-07-31T19:16:53Z)
• The Consistency Dilemma in LLMs: Generator-Evaluator Agreement and Vulnerability to Mistakes [0.0]
  Large language models are increasingly deployed in agentic pipelines that depend on the model evaluating its own outputs without external verification.<n>We propose a new measure, generator-evaluator self-consistency, to test this assumption directly and apply it to 10 frontier models across 491 concepts.
  arXiv  Detail & Related papers  (2026-06-16T03:39:01Z)
• Sensitivity Uncertainty Alignment in Large Language Models [0.0]
  We propose a framework for analyzing failures of large language models under adversarial and ambiguous inputs.<n>We define a scalar score, SUA_theta(x), capturing the difference between distributional sensitivity and predictive entropy.<n>We introduce SUA-TR, a training method combining consistency regularization and entropy alignment, along with an abstention rule for safer inference.
  arXiv  Detail & Related papers  (2026-04-21T17:53:12Z)
• Great Models Think Alike and this Undermines AI Oversight [47.7725284401918]
  We study how model similarity affects both aspects of AI oversight.<n>We propose CAPA: a metric for LM similarity based on overlap in model mistakes.<n>Our work underscores the importance of reporting and correcting for model similarity.
  arXiv  Detail & Related papers  (2025-02-06T18:56:01Z)
• Rethinking LLM Uncertainty: A Multi-Agent Approach to Estimating Black-Box Model Uncertainty [47.95943057892318]
  Quantifying uncertainty in black-box LLMs is vital for reliable responses and scalable oversight.<n>We introduce DiverseAgentEntropy, a novel, theoretically-grounded method employing multi-agent interaction for uncertainty estimation.
  arXiv  Detail & Related papers  (2024-12-12T18:52:40Z)
• CONTESTS: a Framework for Consistency Testing of Span Probabilities in Language Models [16.436592723426305]
  It is unclear whether language models produce the same value for different ways of assigning joint probabilities to word spans. Our work introduces a novel framework, ConTestS, involving statistical tests to assess score consistency across interchangeable completion and conditioning orders.
  arXiv  Detail & Related papers  (2024-09-30T06:24:43Z)
• Uncertainty in Language Models: Assessment through Rank-Calibration [65.10149293133846]
  Language Models (LMs) have shown promising performance in natural language generation. It is crucial to correctly quantify their uncertainty in responding to given inputs. We develop a novel and practical framework, termed $Rank$-$Calibration$, to assess uncertainty and confidence measures for LMs.
  arXiv  Detail & Related papers  (2024-04-04T02:31:05Z)
• When Does Confidence-Based Cascade Deferral Suffice? [69.28314307469381]
  Cascades are a classical strategy to enable inference cost to vary adaptively across samples. A deferral rule determines whether to invoke the next classifier in the sequence, or to terminate prediction. Despite being oblivious to the structure of the cascade, confidence-based deferral often works remarkably well in practice.
  arXiv  Detail & Related papers  (2023-07-06T04:13:57Z)
• Confidence-Based Model Selection: When to Take Shortcuts for Subpopulation Shifts [119.22672589020394]
  We propose COnfidence-baSed MOdel Selection (CosMoS), where model confidence can effectively guide model selection. We evaluate CosMoS on four datasets with spurious correlations, each with multiple test sets with varying levels of data distribution shift.
  arXiv  Detail & Related papers  (2023-06-19T18:48:15Z)
• Underspecification in Language Modeling Tasks: A Causality-Informed Study of Gendered Pronoun Resolution [0.0]
  We introduce a simple causal mechanism to describe the role underspecification plays in the generation of spurious correlations. Despite its simplicity, our causal model directly informs the development of two lightweight black-box evaluation methods.
  arXiv  Detail & Related papers  (2022-09-30T23:10:11Z)
• Cautious Learning of Multiattribute Preferences [2.6151761714896122]
  This paper is dedicated to a cautious learning methodology for predicting preferences between alternatives characterized by binary attributes. By "cautious", we mean that the model learned to represent the multi-attribute preferences is general enough to be compatible with any strict weak order on the alternatives.
  arXiv  Detail & Related papers  (2022-06-15T07:54:16Z)
• How Can We Know When Language Models Know? On the Calibration of Language Models for Question Answering [80.82194311274694]
  We examine the question "how can we know when language models know, with confidence, the answer to a particular query?" We examine three strong generative models -- T5, BART, and GPT-2 -- and study whether their probabilities on QA tasks are well calibrated. We then examine methods to calibrate such models to make their confidence scores correlate better with the likelihood of correctness.
  arXiv  Detail & Related papers  (2020-12-02T03:53:13Z)
• AvgOut: A Simple Output-Probability Measure to Eliminate Dull Responses [97.50616524350123]
  We build dialogue models that are dynamically aware of what utterances or tokens are dull without any feature-engineering. The first model, MinAvgOut, directly maximizes the diversity score through the output distributions of each batch. The second model, Label Fine-Tuning (LFT), prepends to the source sequence a label continuously scaled by the diversity score to control the diversity level. The third model, RL, adopts Reinforcement Learning and treats the diversity score as a reward signal.
  arXiv  Detail & Related papers  (2020-01-15T18:32:06Z)

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.