Position: Semantic Uncertainty Measures Disagreement, Not Reliability

Joseph Hoche    Maxime Corlay    David Brellmann    Andrei Bursuc    Pavel Izmailov    Angela Yao    Gianni Franchi

NeurIPS 2026

Sources of uncertainty in LLMs and LVLMs, from semantic ambiguity to decoding strategies.

Abstract

Large Language Models and their multi-modal variants see increasingly rapid adoption and deployment, including in settings where reliability matters. However their uncertainty remains difficult to assess as classic or early token-level approaches cannot deal with the open-ended outputs of such models. Semantic uncertainty arises as a promising solution to the limits of token-level confidence by measuring disagreement among multiple generated responses. This position paper argues that this framing is incomplete: semantic uncertainty measures semantic disagreement, not actual reliability. A model can be uncertain while producing several valid answers, or certain while repeatedly producing the same wrong answer. We introduce a semantic bias-uncertainty decomposition to show that reliability depends both on variability across meanings and systematic deviation from correct or grounded meanings. This perspective reveals that common hallucination-detection evaluations conflate uncertainty with error. We argue for reliability-centered evaluation that separates semantic disagreement, correctness, and enables a more fine-grained characterization of uncertainty beyond the level of the full answer.

BibTeX

@inproceedings{hoche2026semanticuncertainty,
  title     = {Position: Semantic Uncertainty Measures Disagreement, Not Reliability},
  author    = {Hoche, Joseph and Corlay, Maxime and Brellmann, David and Bursuc, Andrei and Izmailov, Pavel and Yao, Angela and Franchi, Gianni},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year      = {2026}
}