mlsanityData to 7 Oct 2026

The Illusion of Certainty: Uncertainty Quantification for LLMs Fails under Ambiguity

arXiv 2511.04418 ↗First posted 6 Nov 2025cs.LG cs.CLcited by 15

Not on the map. Why?

Abstract (arXiv)

Accurate uncertainty quantification (UQ) in Large Language Models (LLMs) is critical for trustworthy deployment. While real-world language is inherently ambiguous, reflecting aleatoric uncertainty, existing UQ methods are typically benchmarked against tasks with no ambiguity. In this work, we demonstrate that while current uncertainty estimators perform well under the restrictive assumption of no ambiguity, they degrade to close-to-random performance on ambiguous data. To this end, we introduce MAQA* and AmbigQA*, the first ambiguous question-answering (QA) datasets equipped with ground-truth answer distributions estimated from factual co-occurrence. We find this performance deterioration to be consistent across different estimation paradigms: using the predictive distribution itself, internal representations throughout the model, and an ensemble of models. We show that this phenomenon can be theoretically explained, revealing that predictive-distribution and ensemble-based estimators are fundamentally limited under ambiguity. Overall, our study reveals a key shortcoming of current UQ methods for LLMs and motivates a rethinking of current modeling paradigms.

Key facts. Cited by 15 papers in the map (data through 7 Oct 2026). Builds on 35 of its 52 references that are in the map. First posted on arXiv 6 Nov 2025.

Builds on

35 of 52 references.

References · 35

  1. 2016Simple and Scalable Predictive Uncertainty Estimation using Deep Ensemblescited by 3,888
  2. 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,895
  3. 2020The Pile: An 800GB Dataset of Diverse Text for Language Modelingcited by 2,346
  4. 2020DeBERTa: Decoding-enhanced BERT with Disentangled Attentioncited by 2,157
  5. 2023Inference-Time Intervention: Eliciting Truthful Answers from a Language Modelcited by 1,259
  6. 2022When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memoriescited by 1,103
  7. 2019Aleatoric and Epistemic Uncertainty in Machine Learning: An Introduction to Concepts and Methodscited by 903
  8. 2017Deep Bayesian Active Learning with Image Datacited by 812
  9. 2023Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generationcited by 795
  10. 2023Large Language Modelscited by 774
  11. 2018Evidential Deep Learning to Quantify Classification Uncertaintycited by 728
  12. 2021A Survey of Uncertainty in Deep Neural Networkscited by 675
23 more
  1. 2020AmbigQA: Answering Ambiguous Open-domain Questionscited by 385
  2. 2023Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflictscited by 333
  3. 2024INSIDE: LLMs' Internal States Retain the Power of Hallucination Detectioncited by 317
  4. 2024LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinationscited by 230
  5. 2024RedPajama: an Open Dataset for Training Large Language Modelscited by 178
  6. 2024Kernel Language Entropy: Fine-grained Uncertainty Quantification for LLMs from Semantic Similaritiescited by 156
  7. 2024Infini-gram: Scaling Unbounded n-gram Language Models to a Trillion Tokenscited by 148
  8. 2022Quantifying Aleatoric and Epistemic Uncertainty in Machine Learning: Are Conditional Entropy and Mutual Information Appropriate Measures?cited by 124
  9. 2023Decomposing Uncertainty for Large Language Models through Input Clarification Ensemblingcited by 113
  10. 2024Benchmarking Uncertainty Quantification Methods for Large Language Models with LM-Polygraphcited by 105
  11. 2025Uncertainty Quantification and Confidence Calibration in Large Language Models: A Surveycited by 93
  12. 2024Generalization v.s. Memorization: Tracing Language Models' Capabilities Back to Pretraining Datacited by 83
  13. 2024Rethinking Aleatoric and Epistemic Uncertaintycited by 40
  14. 2024Uncertainty quantification in fine-tuned LLMs using LoRA ensemblescited by 35
  15. 2024From Risk to Uncertainty: Generating Predictive Uncertainty Measures via Bayesian Estimationcited by 27
  16. 2025Position: Uncertainty Quantification Needs Reassessment for Large-language Model Agentscited by 27
  17. 2024MAQA: Evaluating Uncertainty Quantification in LLMs Regarding Data Uncertaintycited by 25
  18. 2024Evaluating language models as risk scorescited by 24
  19. 2025From Calibration to Collaboration: LLM Uncertainty Quantification Should Be More Human-Centeredcited by 21
  20. 2024Experts Don't Cheat: Learning What You Don't Know By Predicting Pairscited by 18
  21. 2024Rethinking Uncertainty Estimation in LLMs: A Principled Single-Sequence Measurecited by 18
  22. 2024Provable Uncertainty Decomposition via Higher-Order Calibrationcited by 8
  23. 2025Fine-Grained Uncertainty Decomposition in Large Language Models: A Spectral Approachcited by 8

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 1; 2026: 14.

15 papers.

Sort:

By area:

  1. 2026Uncertainty Quantification in LLM Agents: Foundations, Emerging Challenges, and Opportunitiescited by 15
  2. 2026Verbalizing LLM's Higher-order Uncertainty via Imprecise Probabilitiescited by 4
  3. 2026Mind the Ambiguity: Aleatoric Uncertainty Quantification in LLMs for Safe Medical Question Answeringcited by 2
  4. 2026Task-Awareness Improves LLM Generations and Uncertaintycited by 2
  5. 2026Interpretable Uncertainty for Adaptive Retrieval and Reasoning in Question Answeringcited by 2
  6. 2026AP-OOD: Attention Pooling for Out-of-Distribution Detectioncited by 1
  7. 2025Agile Deliberation: Concept Deliberation for Subjective Visual Classificationcited by 0
  8. 2026Structurally Separated Uncertainty in Supervised Latent Variable Modelscited by 0
  9. 2026Towards Reliable Truth-Aligned Uncertainty Estimation in Large Language Modelscited by 0
  10. 2026When Do We Need LLMs? A Diagnostic for Language-Driven Banditscited by 0
  11. 2026Why Don't You Know? Evaluating the Impact of Uncertainty Sources on Uncertainty Quantification in LLMscited by 0
  12. 2026Evaluating the Relevance of Uncertainty Estimators for LLM Hallucinationcited by 0
3 more
  1. 2026The Role of Ambiguity in Error Prediction via Uncertainty Quantificationcited by 0
  2. 2026Improving LLMs via Validator-to-Generator Alignmentcited by 0
  3. 2026Uncertainty Quantification for LLM Agents: A Taxonomy, an Evaluation Protocol, and an Empirical Studycited by 0