Newest papers

Reasoning-Aware Multicalibrated Reliability Predictors for Language Models

Ekaterina Fedorova, Alessandro Morosini, Chara Podimata · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26645 · cited by 0 works (OpenAlex) · open-access PDF

French

Comme les gens font de plus en plus confiance aux grands modèles de langue (LLM) pour l'information, les fournisseurs de modèles ont besoin de communiquer des estimations significatives de la façon dont les réponses individuelles sont correctes ou, plus généralement, fiables. Dans le contexte des hallucinations de raison, le travail existant utilise la multicibration pour rendre ces prédictions de fiabilité significatives à travers des groupes pertinents pour l'application, tels que les sujets prompt. Les groupes de sujets seuls peuvent cependant masquer des erreurs systématiques: au sein du même sujet et du niveau de confiance, les réponses avec différentes propriétés de chaîne de pensée (CoT) peuvent afficher des erreurs opposées qui annulent la moyenne. Dans ce travail, nous introduis

Machine translation (Meta M2M100); the English original follows.

Abstract

As people increasingly rely on large language models (LLMs) for information, model providers need to communicate meaningful estimates of whether individual responses are correct or, more generally, reliable. In the context of hallucinations, existing work uses multicalibration to make such reliability predictions meaningful across application-relevant groups, such as prompt topics. Topic groups alone, however, can conceal systematic errors: within the same topic and confidence level, responses with different chain-of-thought (CoT) properties may exhibit opposing errors that cancel on average. In this work, we introduce reasoning-aware multicalibration, a framework that supplements application-relevant topic groups with groups derived from observable properties of CoT traces. We highlight that, theoretically, groupings based on CoT are "internally important", i.e., groupings for which the multicalibration guarantee is not societally crucial as it is for high-risk topics, but whose inclusion in the multicalibration procedure could easily and significantly improve predictive accuracy of the final reliability predictor. Empirically, we apply our framework to hallucination prediction and unsafe-response detection across a variety of benchmarks, and find that adding CoT groups to topic-based calibration reduces Brier score in both tasks. Interestingly, we find that multicalibrating with respect to CoT groups alone (without any topic information) already improves calibration across topics, suggesting that the predictor's errors depend on how the model reasoned, rather than what it was asked about in the prompt.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Reliability (semiconductor), Variety (cybernetics), Computer science, Context (archaeology), Calibration, Machine learning

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange