To Think or Not to Think: Exploring the Unthinking Vulnerability in Large Reasoning Models

Zihao Zhu, Hongbao ZHANG, Ruotong Wang, Ke Xu, Siwei Lyu, Baoyuan Wu

Keywords: large reasoning models, chain-of-thought, backdoor attack, adversarial attack, safety alignment, trustworthy AI

TL;DR: We reveal an unthinking vulnerability in Large Reasoning Models and show that it can be exploited from both malicious and beneficial perspectives.

Abstract: Large Reasoning Models (LRMs) are designed to solve complex tasks by generating explicit reasoning traces before producing final answers. However, we reveal a critical vulnerability in LRMs -- termed Unthinking Vulnerability -- wherein the thinking process can be bypassed by manipulating special delimiter tokens. It is empirically demonstrated to be widespread across mainstream LRMs, posing both a significant risk and potential utility, depending on how it is exploited. In this paper, we systematically investigate this vulnerability from both malicious and beneficial perspectives. On the malicious side, we introduce Breaking of Thought (BoT), a novel attack that enables adversaries to bypass the thinking process of LRMs, thereby compromising their reliability and availability. We present two variants of BoT: a training-based version that injects backdoor during the fine-tuning stage, and a training-free version based on adversarial attack during the inference stage. As a potential defense, we propose thinking recovery alignment to partially mitigate the vulnerability. On the beneficial side, we introduce Monitoring of Thought (MoT), a plug-and-play framework that allows model owners to enhance efficiency and safety. It is implemented by leveraging the same vulnerability to dynamically terminate redundant or risky reasoning through external monitoring. Extensive experiments show that BoT poses a significant threat to reasoning reliability, while MoT provides a practical solution for preventing overthinking and jailbreaking. Our findings expose an inherent flaw in current LRM architectures and underscore the need for more robust reasoning systems in the future. Code is available at https://anonymous.4open.science/r/unthinking_vulnerability.

Submission Number: 69