Wait, We Don't Need to "Wait"! Removing Thinking Tokens Improves Reasoning Efficiency
shuaishuaicdp
Abstract
NoWait suppresses explicit self-reflection tokens during inference to enhance efficiency in multimodal reasoning without reducing model utility.
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.
Community
🚀 Do we really need to "Wait" in AI reasoning?
NEW RESEARCH: Removing "Wait", "Hmm" thinking tokens BOOSTS efficiency by 27%-51%! 🤯
🔥 Key Findings
❌ "Wait, let me think again..."
❌ "Hmm, maybe I should..."
✅ Direct reasoning = 2x efficiency!
⚡ NoWait Method Highlights:
🎯 Training-Free: Plug-and-play solution
📊 Massive Token Reduction: Up to 51% shorter outputs
🎯 Accuracy Preserved: Performance maintained or improved
🌐 Multimodal: Text + Vision + Video reasoning
📈 Extensive Validation:
• 10 benchmarks tested
• 5 R1-style model families
• QwQ-32B, Phi4, Qwen3, Kimi-VL, QvQ models
💡 Core Insight:
Explicit self-reflection ≠ Better reasoning
Simple keyword suppression → Dramatic efficiency gains
This could reshape how we think about AI reasoning! 🤖✨
Paper: https://arxiv.org/pdf/2506.08343
#AI #MachineLearning #Reasoning #Efficiency #LLM #Research
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- Dynamic Early Exit in Reasoning Models (2025)
- DRP: Distilled Reasoning Pruning with Skill-aware Step Decomposition for Efficient Large Reasoning Models (2025)
- When to Continue Thinking: Adaptive Thinking Mode Switching for Efficient Reasoning (2025)
- ThinkSwitcher: When to Think Hard, When to Think Fast (2025)
- ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoning (2025)
- ThinkLess: A Training-Free Inference-Efficient Method for Reducing Reasoning Redundancy (2025)
- Done Is Better than Perfect: Unlocking Efficient Reasoning by Structured Multi-Turn Decomposition (2025)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment:
@librarian-bot
recommend
Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper