Wait, We Don't Need to "Wait"! Removing Thinking Tokens Improves Reasoning Efficiency
Chenlong Wang, Yuanning Feng, Dongping Chen, Zhaoyang Chu, Ranjay Krishna, Tianyi Zhou
TL;DR
The paper investigates whether explicit self-reflection tokens are necessary for reasoning in large reasoning models and introduces NoWait, an inference-time logit-suppression technique that disables reflection keywords during decoding. By pruning tokens such as Wait and Hmm , NoWait reduces chain-of-thought length by up to 51% across textual, visual, and video reasoning without significantly harming accuracy, and it operates as a plug-and-play solution across multiple model families. Extensive experiments across ten benchmarks demonstrate consistent efficiency gains across RL-based and multimodal models, while revealing that distilled models are more sensitive to the removal of reflection cues. The work provides practical guidance for deploying efficient reasoning systems and suggests that explicit waiting-like reflections are not essential for achieving high-quality reasoning in modern LRMs.
Abstract
Recent advances in large reasoning models have enabled complex, step-by-step reasoning but often introduce significant overthinking, resulting in verbose and redundant outputs that hinder efficiency. In this study, we examine whether explicit self-reflection, signaled by tokens such as "Wait" and "Hmm", is necessary for advanced reasoning. We propose NoWait, a simple yet effective approach that disables explicit self-reflection by suppressing these tokens during inference. Extensive experiments on ten benchmarks across textual, visual, and video reasoning tasks show that NoWait reduces chain-of-thought trajectory length by up to 27%-51% in five R1-style model series, without compromising model utility. NoWait thus offers a plug-and-play solution for efficient and utility-preserving multimodal reasoning.