# Same Answer, Different Confidence: Protocol Sensitivity in LLM Confidence Calibration

arXiv: https://arxiv.org/abs/2605.27752 · HTML: https://mlsanity.com/paper/2605.27752/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Builds on 30 of its 36 references that are in the map.
- First posted on arXiv 26 May 2026.

## Builds on (most cited first, 30 of 30)

- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension](https://mlsanity.com/paper/1705.03551/) (2017)
- [Calibrate Before Use: Improving Few-Shot Performance of Language Models](https://mlsanity.com/paper/2102.09690/) (2021)
- [Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs](https://mlsanity.com/paper/2306.13063/) (2023)
- [Just Ask for Calibration: Strategies for Eliciting Calibrated Confidence Scores from Language Models Fine-Tuned with Human Feedback](https://mlsanity.com/paper/2305.14975/) (2023)
- [Crowdsourcing Multiple Choice Science Questions](https://mlsanity.com/paper/1707.06209/) (2017)
- [Teaching Models to Express Their Uncertainty in Words](https://mlsanity.com/paper/2205.14334/) (2022)
- [Measuring Calibration in Deep Learning](https://mlsanity.com/paper/1904.01685/) (2019)
- [Calibration of Pre-trained Transformers](https://mlsanity.com/paper/2003.07892/) (2020)
- [Verified Uncertainty Calibration](https://mlsanity.com/paper/1909.10155/) (2019)
- [Reducing conversational agents' overconfidence through linguistic calibration](https://mlsanity.com/paper/2012.14983/) (2020)
- [Mix-n-Match: Ensemble and Compositional Methods for Uncertainty Calibration in Deep Learning](https://mlsanity.com/paper/2003.07329/) (2020)
- [Mitigating Bias in Calibration Error Estimation](https://mlsanity.com/paper/2012.08668/) (2020)
- [On Verbalized Confidence Scores for LLMs](https://mlsanity.com/paper/2412.14737/) (2024)
- [Calibrating Large Language Models with Sample Consistency](https://mlsanity.com/paper/2402.13904/) (2024)
- [Reasoning Models Better Express Their Confidence](https://mlsanity.com/paper/2505.14489/) (2025)
- [Smooth ECE: Principled Reliability Diagrams via Kernel Smoothing](https://mlsanity.com/paper/2309.12236/) (2023)
- [Calibrating Verbal Uncertainty as a Linear Feature to Reduce Hallucinations](https://mlsanity.com/paper/2503.14477/) (2025)
- [Distribution-free calibration guarantees for histogram binning without sample splitting](https://mlsanity.com/paper/2105.04656/) (2021)
- [Uncertainty in Language Models: Assessment through Rank-Calibration](https://mlsanity.com/paper/2404.03163/) (2024)
- [Restoring Calibration for Aligned Large Language Models: A Calibration-Aware Fine-Tuning Approach](https://mlsanity.com/paper/2505.01997/) (2025)
- [Are Large Language Models More Honest in Their Probabilistic or Verbalized Confidence?](https://mlsanity.com/paper/2408.09773/) (2024)
- [Calibrating LLM Confidence by Probing Perturbed Representation Stability](https://mlsanity.com/paper/2505.21772/) (2025)
- [Your Pre-trained LLM is Secretly an Unsupervised Confidence Calibrator](https://mlsanity.com/paper/2505.16690/) (2025)
- [Thinking Out Loud: Do Reasoning Models Know When They're Right?](https://mlsanity.com/paper/2504.06564/) (2025)
- [BAS: A Decision-Theoretic Approach to Evaluating Large Language Model Confidence](https://mlsanity.com/paper/2604.03216/) (2026)
- [Calibration Is Not Enough: Evaluating Confidence Estimation Under Language Variations](https://mlsanity.com/paper/2601.08064/) (2026)
- [Semantic Calibration Prevails Where Token Confidence Fails: Benchmarking Long-Form Scientific QA](https://mlsanity.com/paper/2602.00279/) (2026)

## Built on by (most cited first, 6 of 6)

- [Probability Contracts: Accuracy, Coherence, and Decisions Across LLM Interfaces](https://mlsanity.com/paper/2609.37470/) (2026)
- [When Policies Change Probabilities: Modular Decision-Making for LLM Code Review](https://mlsanity.com/paper/2608.02677/) (2026)
- [Confident at the moment of action: belief miscalibration in LLM play under hidden information](https://mlsanity.com/paper/2608.24691/) (2026)
- [Uncertainty Quantification for LLM Agents: A Taxonomy, an Evaluation Protocol, and an Empirical Study](https://mlsanity.com/paper/2609.07395/) (2026)
- [Scored vs. Generated Readouts in Behavioral Language Models: An Empirical Study of Elicitation Format](https://mlsanity.com/paper/2609.09882/) (2026)
- [Replication Without Persistence in Hosted LLMs: Measurement Sensitivity in Action-Time Belief Evaluation](https://mlsanity.com/paper/2609.22478/) (2026)

## Abstract (arXiv)

Is verbalized confidence better calibrated than token likelihood? The answer depends on how the token likelihood is measured: which answer is scored, and under which prompt. Published comparisons diverge on this, and in a twelve-study audit five never state the choice. We fix one prediction event per question, the model's own answer together with its correctness label, and score that same answer under a plain query and inside the confidence prompt, holding the answer and its label fixed. Across four QA datasets and three 7-8B Instruct models this changes which signal performs better, by point estimate, in 4 of 12 settings under ECE and 9 of 12 under AUROC. The AUROC result cannot come from rescaling the likelihoods, since AUROC is invariant to any common order-preserving transformation; the items are ordered differently. Two further choices behave the same way: substituting the reference string for the model's own answer, and reading the first answer token instead of the answer span. Crossing three answer slots, two scored strings, and two readouts gives twelve measured operational variants that leave the sign of the ECE comparison ndetermined in 6 of 12 settings, whereas alternative calibration estimators move it substantially less, although one changes a single prompted-context winner. Verbalized confidence is sensitive to answer formulation as well: replacing an accepted TriviaQA alias with the canonical reference raises confidence by 0.072 although both answers are correct. Comparing the two signals therefore requires an explicit answer, context, and evaluation protocol.