The reviewed record of science sign in
Pith

arxiv: 2503.14477 · v2 · pith:SSMJYBT4 · submitted 2025-03-18 · cs.CL

Calibrating Verbal Uncertainty as a Linear Feature to Reduce Hallucinations

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:SSMJYBT4record.jsonopen to challenge →

classification cs.CL
keywords uncertaintyhallucinationsverbalsemanticachievingactualfeaturelanguage
0
0 comments X
read the original abstract

LLMs often adopt an assertive language style also when making false claims. Such ``overconfident hallucinations'' mislead users and erode trust. Achieving the ability to express in language the actual degree of uncertainty around a claim is therefore of great importance. We find that ``verbal uncertainty'' is governed by a single linear feature in the representation space of LLMs, and show that this has only moderate correlation with the actual ``semantic uncertainty'' of the model. We apply this insight and show that (1) the mismatch between semantic and verbal uncertainty is a better predictor of hallucinations than semantic uncertainty alone and (2) we can intervene on verbal uncertainty at inference time and reduce confident hallucinations on short-form answers, achieving an average relative reduction of ~30%.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Can LLMs Use Linguistic Uncertainty Markers to Reliably Reflect Intrinsic Confidence?

    cs.CL 2026-05 unverdicted novelty 7.0

    LLMs struggle to associate epistemic markers with stable internal confidence levels across distributions, even under model-centric interpretations, while maintaining somewhat consistent marker rankings.

  2. Explaining Sources of Uncertainty in Automated Fact-Checking

    cs.CL 2025-05 unverdicted novelty 7.0

    CLUE generates natural language explanations of model uncertainty in fact-checking by unsupervised identification of claim-evidence and inter-evidence conflicts and agreements, followed by prompting and attention steering.

  3. Estimating Uncertainty from Reasoning: A Large-Scale Study of Multi- and Crosslingual MCQA Performance in LLMs

    cs.CL 2026-07 conditional novelty 6.0

    A large-scale multilingual evaluation of LLM uncertainty estimation methods across 22 languages and 9 models finds that English reasoning closes the UE gap for low-resource languages and that optimal UE method choice ...

  4. Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs

    cs.CL 2026-06 unverdicted novelty 6.0

    RLMF uses quality of model self-judgments to refine RL rankings and select training data, achieving SOTA faithful calibration while preserving accuracy and outperforming standard RL by up to 63%.

  5. Quantifying Faithful Confidence Expression in Large Reasoning Models

    cs.CL 2026-06 unverdicted novelty 6.0

    A new framework quantifies faithful confidence expression in large reasoning models by comparing linguistic decisiveness to token probabilities, hidden states, and response consistency, revealing it as a persistent challenge.

  6. Hallucinations Undermine Trust; Metacognition is a Way Forward

    cs.CL 2026-05 unverdicted novelty 6.0

    LLMs need metacognition to align expressed uncertainty with their actual knowledge boundaries, moving beyond knowledge expansion to reduce confident errors.

  7. GrACE: A Generative Approach to Better Confidence Elicitation and Efficient Test-Time Scaling in Large Language Models

    cs.CL 2025-09 unverdicted novelty 6.0

    GrACE is a fine-tuned generative method that uses similarity to a special token embedding for real-time calibrated confidence in LLMs and enables efficient confidence-based test-time scaling.

  8. SEAT: Sparse Entity-Aware Tuning for Knowledge Adaptation while Preserving Epistemic Abstention

    cs.AI 2025-06 unverdicted novelty 6.0

    SEAT preserves epistemic abstention in LLMs during knowledge adaptation via sparse tuning and entity-perturbed KL regularization, yielding 18-101% better abstention on unknown queries while retaining near-perfect know...