Can LLMs Use Linguistic Uncertainty Markers to Reliably Reflect Intrinsic Confidence?
Not on the map. Why?
Abstract (arXiv)
LLMs' linguistically expressed confidence should faithfully reflect their intrinsic uncertainty. While recent work shows LLMs struggle to use epistemic markers (e.g., "it is likely...") in a human-aligned fashion, it remains unclear whether models can apply their own linguistic confidence framework to associate markers with specific confidence levels in a stable and generalizable way, and how contextual features impact this ability. We conduct the first systematic study of this question, formalizing _marker internal confidence_ (MIC) as the estimated intrinsic confidence a model associates with a specific epistemic marker in a given task domain. We present 7 metrics to evaluate the stability of MICs within and across distributions. Applying our analysis framework to diverse models and tasks, we find that LLMs remain faithfully miscalibrated even under model-centric interpretation of marker meanings, struggling to differentiate markers by internal confidence across distributions despite preserving a somewhat consistent ranking order across tasks. This supplies critical, complementary evidence to existing work toward a holistic understanding of faithful calibration in LLMs, emphasizing the need for more aligned and stable marker use to improve trustworthiness and reliability.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 64 of its 96 references that are in the map. First posted on arXiv 27 May 2026.
Builds on
64 of 96 references.
References · 64
- 2020Measuring Massive Multitask Language Understandingcited by 7,768
- 2018GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understandingcited by 4,854
- 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,710
- 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,919
- 2019SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systemscited by 1,718
- 2023A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questionscited by 1,582
- 2023Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMscited by 946
- 2023Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generationcited by 795
- 2023Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedbackcited by 788
- 2023SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Modelscited by 787
- 2017Crowdsourcing Multiple Choice Science Questionscited by 742
- 2022Discovering Latent Knowledge in Language Models Without Supervisioncited by 733
38 more
- 2022Teaching Models to Express Their Uncertainty in Wordscited by 731
- 2023The Internal State of an LLM Knows When It's Lyingcited by 644
- 2020AmbigQA: Answering Ambiguous Open-domain Questionscited by 385
- 2024Measuring short-form factuality in large language modelscited by 345
- 2023HaluEval: A Large-Scale Hallucination Evaluation Benchmark for Large Language Modelscited by 329
- 2020Effect of Confidence and Explanation on Accuracy and Trust Calibration in AI-Assisted Decision Makingcited by 281
- 2024A Comprehensive Survey of Hallucination Mitigation Techniques in Large Language Modelscited by 266
- 2023Do Large Language Models Know What They Don't Know?cited by 238
- 2023Fine-tuning Language Models for Factualitycited by 219
- 2020Reducing conversational agents' overconfidence through linguistic calibrationcited by 209
- 2021On Hallucination and Predictive Uncertainty in Conditional Language Generationcited by 176
- 2024Kernel Language Entropy: Fine-grained Uncertainty Quantification for LLMs from Semantic Similaritiescited by 156
- 2023Shifting Attention to Relevance: Towards the Predictive Uncertainty Quantification of Free-Form Large Language Modelscited by 147
- 2023Quantifying Uncertainty in Answers from any Language Model and Enhancing their Trustworthinesscited by 121
- 2024To Believe or Not to Believe Your LLMcited by 120
- 2023Decomposing Uncertainty for Large Language Models through Input Clarification Ensemblingcited by 113
- 2024What Large Language Models Know and What People Think They Knowcited by 109
- 2023R-Tuning: Instructing Large Language Models to Say `I Don't Know'cited by 106
- 2024Relying on the Unreliable: The Impact of Language Models' Reluctance to Express Uncertaintycited by 106
- 2023Navigating the Grey Area: How Expressions of Uncertainty and Overconfidence Affect Language Modelscited by 91
- 2024Can Large Language Models Faithfully Express Their Intrinsic Uncertainty in Words?cited by 85
- 2024On Verbalized Confidence Scores for LLMscited by 83
- 2024Linguistic Calibration of Long-Form Generationscited by 59
- 2025A Survey of Uncertainty Estimation Methods on Large Language Modelscited by 57
- 2024LegalAgentBench: Evaluating LLM Agents in Legal Domaincited by 56
- 2024Large Language Models for Disease Diagnosis: A Scoping Reviewcited by 51
- 2025Calibrating Verbal Uncertainty as a Linear Feature to Reduce Hallucinationscited by 48
- 2023Look Before You Leap: An Exploratory Study of Uncertainty Measurement for Large Language Modelscited by 40
- 2023Llamas Know What GPTs Don't Show: Surrogate Models for Confidence Estimationcited by 38
- 2024Thermometer: Towards Universal Calibration for Large Language Modelscited by 38
- 2025Evaluating Large Language Models in Scientific Discoverycited by 38
- 2023Quantifying Uncertainty in Natural Language Explanations of Large Language Modelscited by 31
- 2024Improving Uncertainty Quantification in Large Language Models via Semantic Embeddingscited by 30
- 2024Are LLM-Judges Robust to Expressions of Uncertainty? Investigating the effect of Epistemic Markers on LLM-based Evaluationcited by 29
- 2024Fact-and-Reflection (FaR) Improves Confidence Calibration of Large Language Modelscited by 27
- 2024Cycles of Thought: Measuring LLM Confidence through Stable Explanationscited by 27
- 2024Don't Go To Extremes: Revealing the Excessive Sensitivity and Calibration Limitations of LLMs in Implicit Hate Speech Detectioncited by 25
- 2024Finetuning Language Models to Emit Linguistic Expressions of Uncertaintycited by 25
50 most cited of 64 All 64 (JSON)
Built on by
Citing papers per year: 2026: 2.
2 papers.