Adaptive Conformal Prediction for Improving Factuality of Generations by Large Language Models

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Rubashevskii, Aleksandr, Piatrashyn, Dzianis, Nakov, Preslav, Panov, Maxim
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913035463950336
author Rubashevskii, Aleksandr
Piatrashyn, Dzianis
Nakov, Preslav
Panov, Maxim
author_facet Rubashevskii, Aleksandr
Piatrashyn, Dzianis
Nakov, Preslav
Panov, Maxim
contents Large language models (LLMs) are prone to generating factually incorrect outputs. Recent work has applied conformal prediction to provide uncertainty estimates and statistical guarantees for the factuality of LLM generations. However, existing approaches are typically not prompt-adaptive, limiting their ability to capture input-dependent variability. As a result, they may filter out too few items (leading to over-coverage) or too many (under-coverage) for a given task or prompt. We propose an adaptive conformal prediction approach that extends conformal score transformation methods to LLMs, with applications to long-form generation and multiple-choice question answering. This enables prompt-dependent calibration, retaining marginal coverage guarantees while improving conditional coverage. In addition, the approach naturally supports selective prediction, allowing unreliable claims or answer choices to be filtered out in downstream applications. We evaluate our approach on multiple white-box models across diverse domains and show that it significantly outperforms existing baselines in terms of conditional coverage.
format Preprint
id arxiv_https___arxiv_org_abs_2604_13991
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Adaptive Conformal Prediction for Improving Factuality of Generations by Large Language Models
Rubashevskii, Aleksandr
Piatrashyn, Dzianis
Nakov, Preslav
Panov, Maxim
Computation and Language
Artificial Intelligence
Machine Learning
Large language models (LLMs) are prone to generating factually incorrect outputs. Recent work has applied conformal prediction to provide uncertainty estimates and statistical guarantees for the factuality of LLM generations. However, existing approaches are typically not prompt-adaptive, limiting their ability to capture input-dependent variability. As a result, they may filter out too few items (leading to over-coverage) or too many (under-coverage) for a given task or prompt. We propose an adaptive conformal prediction approach that extends conformal score transformation methods to LLMs, with applications to long-form generation and multiple-choice question answering. This enables prompt-dependent calibration, retaining marginal coverage guarantees while improving conditional coverage. In addition, the approach naturally supports selective prediction, allowing unreliable claims or answer choices to be filtered out in downstream applications. We evaluate our approach on multiple white-box models across diverse domains and show that it significantly outperforms existing baselines in terms of conditional coverage.
title Adaptive Conformal Prediction for Improving Factuality of Generations by Large Language Models
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2604.13991