mlsanityData to 14 Oct 2026

Truth for Believable AI: Expressed Doubt, Provenance, and Belief Revision as an Engineerable Stance

arXiv 2609.26035 ↗First posted 22 Sep 2026cs.CL cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Conversational agents often express answers in a uniformly confident register. We test whether expressed uncertainty, provenance-aware assertion, and explicit belief revision can be implemented as a behavior layer over a fixed language model; we do not test believability or trust. The layer combines three epistemic states, per-claim confidence and typed provenance, a provenance-gated expression rule, and a persistent revision store with auditable acknowledgments and partial resistance to false corrections. We evaluate it on a constructed, mechanically scored multi-session benchmark using a synthetic model and Qwen2.5-0.5B-Instruct. The synthetic instrument passes all five checks. On the real model, acknowledgment soundness, a by-construction guarantee, holds in 100% of cases, and true corrections are accepted more often than false ones (0.44 vs. 0.15 on held beliefs; 0.875 vs. 0.420 including rule-accepted corrections of unheld facts), but the pre-specified expression-fidelity, contradiction-separation, and provenance margins fail. A disclosed post hoc analysis shows that expression gated on mean answer-token probability ranks correctness below chance end to end (AUC 0.41, conversation-clustered), whereas gating on sampling consistency discriminates (AUC 0.66). A consistency-gated configuration selected from this finding and evaluated under a separately committed protocol meets the conversation-level manipulation and capability-equivalence criteria and replicates on a redrawn conversation set. The manipulation result is selection-dependent, and both criteria remain unresolved when uncertainty is clustered over the 60 facts. The supported conclusions are limited to the by-construction audit guarantee, store-dependent partial correction discrimination, and a benchmark- and model-specific failure of token-probability gating; scaling the fact base is required before human evaluation.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 33 of its 44 references that are in the map. First posted on arXiv 22 Sep 2026.

Builds on

33 of 44 references.

References · 33

  1. 2022Locating and Editing Factual Associations in GPTcited by 2,691
  2. 2023Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflectioncited by 1,839
  3. 2018Know What You Don't Know: Unanswerable Questions for SQuADcited by 1,554
  4. 2023Towards Understanding Sycophancy in Language Modelscited by 1,040
  5. 2022Mass-Editing Memory in a Transformercited by 1,001
  6. 2023Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMscited by 974
  7. 2024Evaluating Very Long-Term Conversational Memory of LLM Agentscited by 830
  8. 2023Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generationcited by 819
  9. 2023Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedbackcited by 812
  10. 2023SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Modelscited by 810
  11. 2022Teaching Models to Express Their Uncertainty in Wordscited by 751
  12. 2023Enabling Large Language Models to Generate Text with Citationscited by 595
21 more
  1. 2023Editing Large Language Models: Problems, Methods, and Opportunitiescited by 372
  2. 2023Role-Play with Large Language Modelscited by 294
  3. 2022RARR: Researching and Revising What Language Models Say, Using Language Modelscited by 287
  4. 2021Beyond Goldfish Memory: Long-Term Open-Domain Conversationcited by 267
  5. 2021Measuring Attribution in Natural Language Generation Modelscited by 256
  6. 2020Selective Question Answering under Domain Shiftcited by 238
  7. 2023Evaluating Large Language Models in Theory of Mind Taskscited by 222
  8. 2020Reducing conversational agents' overconfidence through linguistic calibrationcited by 210
  9. 2018Dialogue Natural Language Inferencecited by 135
  10. 2024Know Your Limits: A Survey of Abstention in Large Language Modelscited by 121
  11. 2022Attributed Question Answering: Evaluation and Modeling for Attributed Large Language Modelscited by 112
  12. 2023R-Tuning: Instructing Large Language Models to Say `I Don't Know'cited by 111
  13. 2023Navigating the Grey Area: How Expressions of Uncertainty and Overconfidence Affect Language Modelscited by 91
  14. 2020I like fish, especially dolphins: Addressing Contradictions in Dialogue Modelingcited by 68
  15. 2024"I'm Not Sure, But...": Examining the Impact of Large Language Models' Uncertainty Expression on User Reliance and Trustcited by 67
  16. 2024Linguistic Calibration of Long-Form Generationscited by 60
  17. 2024Belief Revision: The Adaptability of Large Language Models Reasoningcited by 30
  18. 2024Fundamental Problems With Model Editing: How Should Rational Belief Revision Work in LLMs?cited by 24
  19. 2026Belief Memory: Agent Memory Under Partial Observabilitycited by 11
  20. 2026Perceived AGI: Believability as Dimensional Completeness, Not Capabilitycited by 2
  21. 2026Entropy in Conversational AI: Structured Unpredictability as Inferrable Interioritycited by 1

Built on by

Not cited yet.