Paperlane

no. 16 · Tuesday 17 November 2026

PAC Reasoning: Controlling the Performance Loss for Efficient Reasoning

Hao Zeng, Jianguo Huang, Bingyi Jing, Hongxin Wei and Bo An

math.ST · v4 · code

Abstract

Large reasoning models (LRMs) have achieved remarkable progress in complex problem-solving tasks. Despite this success, LRMs typically suffer from high computational costs during deployment, highlighting a need for efficient inference. A practical direction is to switch the LRM between thinking and non-thinking modes dynamically. However, such approaches often introduce additional reasoning errors and lack statistical guarantees for the performance loss, which are critical for high-stakes applications. In this work, we propose Probably Approximately Correct (PAC) Reasoning that controls the performance loss under the user-specified tolerance. Specifically, we construct an upper confidence bound on the performance loss and determine a threshold for switching to the non-thinking model. Theoretically, using the threshold to switch between the thinking and non-thinking modes ensures bounded performance loss in a distribution-free manner. Our experiments on reasoning benchmarks show that the proposed method can save computational budgets and control the user-specified performance loss.

code
github.com/Jianguo99/PAC-Reasoning
comments
Accepted at NeurIPS 2026 (poster). Camera-ready version. Project page: https://zenghao-stat.github.io/pages/pac_reasoning/. Code: https://github.com/Jianguo99/PAC-Reasoning
fields
Statistics Theory
pdf ↗ page arXiv:2510.09133 ↗