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

TL;DR

This paper introduces an adaptive conformal prediction method for large language models that improves factuality and conditional coverage by enabling prompt-dependent calibration and selective prediction.

Contribution

It extends conformal score transformation techniques to LLMs, providing input-aware uncertainty estimates and enhancing factuality filtering capabilities.

Findings

Significantly outperforms existing baselines in conditional coverage.

Supports prompt-dependent calibration for better factuality assessment.

Enables selective prediction to filter unreliable outputs.

Abstract

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.