Quantifying and Understanding Uncertainty in Large Reasoning Models

2026-04-14

AI Summary

This work addresses the limitations of existing large reasoning models, which lack effective quantification of generation uncertainty, fail to model the logical relationship between reasoning trajectories and final answers, and cannot provide statistical guarantees under limited samples. To overcome these challenges, the authors propose a unified framework that integrates conformal prediction with Shapley value analysis, offering—for the first time—a theoretically grounded, distribution-free, and model-agnostic approach that jointly models uncertainty in both the reasoning process and the final answer. The method not only yields high-coverage prediction intervals but also enables precise attribution of key training examples and critical reasoning steps. Extensive experiments across multiple complex reasoning benchmarks demonstrate the framework’s effectiveness and interpretability.

Technology Category

• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty
• Machine Learning: Calibration & Uncertainty Quantification
• Knowledge Representation and Reasoning: Reasoning with Beliefs

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Large Reasoning Models (LRMs) have recently demonstrated significant improvements in complex reasoning. While quantifying generation uncertainty in LRMs is crucial, traditional methods are often insufficient because they do not provide finite-sample guarantees for reasoning-answer generation. Conformal prediction (CP) stands out as a distribution-free and model-agnostic methodology that constructs statistically rigorous uncertainty sets. However, existing CP methods ignore the logical connection between the reasoning trace and the final answer. Additionally, prior studies fail to interpret the origins of uncertainty coverage for LRMs as they typically overlook the specific training factors driving valid reasoning. Notably, it is challenging to disentangle reasoning quality from answer correctness when quantifying uncertainty, while simultaneously establishing theoretical guarantees for computationally efficient explanation methods. To address these challenges, we first propose a novel methodology that quantifies uncertainty in the reasoning-answer structure with statistical guarantees. Subsequently, we develop a unified example-to-step explanation framework using Shapley values that identifies a provably sufficient subset of training examples and their key reasoning steps to preserve the guarantees. We also provide theoretical analyses of our proposed methods. Extensive experiments on challenging reasoning datasets verify the effectiveness of the proposed methods.

Problem

• Large Reasoning Models
• Uncertainty Quantification
• Conformal Prediction
• Reasoning Trace
• Statistical Guarantees

Innovation

• Conformal Prediction
• Large Reasoning Models
• Uncertainty Quantification
• Shapley Values
• Reasoning Trace