mlsanityData to 7 Oct 2026

Shapley Uncertainty in Natural Language Generation

arXiv 2507.21406 ↗First posted 29 Jul 2025cs.AIcited by 2

On the map: LLM alignment / RLHF / preference optimisation → LLM preference optimization methods

Abstract (arXiv)

In question-answering tasks, determining when to trust the outputs is crucial to the alignment of large language models (LLMs). Kuhn et al. (2023) introduces semantic entropy as a measure of uncertainty, by incorporating linguistic invariances from the same meaning. It primarily relies on setting threshold to measure the level of semantic equivalence relation. We propose a more nuanced framework that extends beyond such thresholding by developing a Shapley-based uncertainty metric that captures the continuous nature of semantic relationships. We establish three fundamental properties that characterize valid uncertainty metrics and prove that our Shapley uncertainty satisfies these criteria. Through extensive experiments, we demonstrate that our Shapley uncertainty more accurately predicts LLM performance in question-answering and other datasets, compared to similar baseline measures.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in LLM preference optimization methods, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 29 of its 50 references that are in the map. First posted on arXiv 29 Jul 2025.

Builds on

29 of 50 references.

Closest ancestors · 1

Same direction, newest first

  1. 2023Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedback24 May

Other references · 28

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  3. 2022PaLM: Scaling Language Modeling with Pathwayscited by 5,304
  4. 2016SQuAD: 100,000+ Questions for Machine Comprehension of Textcited by 4,115
  5. 2022OPT: Open Pre-trained Transformer Language Modelscited by 3,464
  6. 2017TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehensioncited by 2,895
  7. 2017What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?cited by 2,335
  8. 2020DeBERTa: Decoding-enhanced BERT with Disentangled Attentioncited by 2,157
  9. 2022Language Models (Mostly) Know What They Knowcited by 1,783
  10. 2021Are NLP Models really able to Solve Simple Math Word Problems?cited by 861
  11. 2020A Review of Uncertainty Quantification in Deep Learning: Techniques, Applications and Challengescited by 813
  12. 2022Teaching Models to Express Their Uncertainty in Wordscited by 731
16 more
  1. 2019Latent Retrieval for Weakly Supervised Open Domain Question Answeringcited by 691
  2. 2018CoQA: A Conversational Question Answering Challengecited by 674
  3. 2020Calibration of Pre-trained Transformerscited by 307
  4. 2021Unsolved Problems in ML Safetycited by 291
  5. 2020Unsupervised Quality Estimation for Neural Machine Translationcited by 210
  6. 2023Conformal Prediction with Large Language Models for Multi-Choice Question Answeringcited by 131
  7. 2024Language Models with Conformal Factuality Guaranteescited by 128
  8. 2024Uncertainty Estimation and Quantification for LLMs: A Simple Supervised Approachcited by 73
  9. 2023Do Androids Know They're Only Dreaming of Electric Sheep?cited by 67
  10. 2024Do LLMs Know about Hallucination? An Empirical Investigation of LLM's Hidden Statescited by 50
  11. 2021Knowing More About Questions Can Help: Improving Calibration in Question Answeringcited by 47
  12. 2022Enhancing Adversarial Training with Second-Order Statistics of Weightscited by 42
  13. 2020How does Weight Correlation Affect the Generalisation Ability of Deep Neural Networkscited by 32
  14. 2021Can Explanations Be Useful for Calibrating Black Box Models?cited by 20
  15. 2023TrajPAC: Towards Robustness Verification of Pedestrian Trajectory Prediction Modelscited by 8
  16. 2023Reducing LLM Hallucinations using Epistemic Neural Networkscited by 8

Built on by

Cited by year (2026 to 7 Oct)

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

2 papers.

  1. 2025Chain-of-Thought as a Lens: Evaluating Structured Reasoning Alignment between Human Preferences and Large Language Modelscited by 5
  2. 2026Beyond Semantic Equivalence: Logical Graphs for LLM Uncertainty Quantificationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024MAQA: Evaluating Uncertainty Quantification in LLMs Regarding Data Uncertainty
  2. 2023Quantifying Uncertainty in Natural Language Explanations of Large Language Models
  3. 2025Efficient semantic uncertainty quantification in language models via diversity-steered sampling
  4. 2024Rethinking Uncertainty Estimation in LLMs: A Principled Single-Sequence Measure
  5. 2026URAG: A Benchmark for Uncertainty Quantification in Retrieval-Augmented Large Language Models
  6. 2025SUGAR: Leveraging Contextual Confidence for Smarter Retrieval
  7. 2025Tools in the Loop: Quantifying Uncertainty of LLM Question Answering Systems That Use Tools
  8. 2026Semantic Uncertainty Quantification Needs Factual Equivalence