Large reasoning models (LRMs) achieve state-of-the-art performance by generating long chains-of-thought, but often waste computation on redundant reasoning after the correct answer has already been reached. We introduce Early-Stopping for Token-Aware Reasoning (ESTAR), which detects and reduces such reasoning redundancy to improve efficiency without sacrificing accuracy. Our method combines (i) a trajectory-based classifier that identifies when reasoning can be safely stopped, (ii) supervised fine-tuning to teach LRMs to propose self-generated <stop> signals, and (iii) <stop>-aware reinforcement learning that truncates rollouts at self-generated stop points with compute-aware rewards. Experiments on four reasoning datasets show that ESTAR reduces reasoning length by about 3.7x (from 4,799 to 1,290) while preserving accuracy (74.9% vs. 74.2%), with strong cross-domain generalization. These results highlight early stopping as a simple yet powerful mechanism for improving reasoning efficiency in LRMs.

翻译：大型推理模型通过生成长链思维实现了最先进的性能，但往往在已经得出正确答案后仍继续冗余推理，从而浪费计算资源。我们提出了面向令牌感知推理的早期停止方法，该方法通过检测并减少此类推理冗余，在不牺牲准确性的前提下提升效率。我们的方法结合了三个关键组件：(i) 基于轨迹的分类器，用于识别何时可安全停止推理；(ii) 监督微调，指导大型推理模型生成自发的<stop>信号；(iii) 具备<stop>感知的强化学习，在自生成停止点处截断推演过程，并采用计算感知的奖励机制。在四个推理数据集上的实验表明，ESTAR 将推理长度缩短约 3.7 倍（从 4,799 降至 1,290），同时保持准确性（74.9% 对比 74.2%），并展现出强大的跨领域泛化能力。这些结果凸显了早期停止作为一种简单而有效的机制，在提升大型推理模型推理效率方面的重要价值。