Efficient Reasoning Through Suppression of Self-Affirmation Reflections in Large Reasoning Models

2025-06-14

AI Summary

Large reasoning models exhibit “self-affirmative reflection”—a phenomenon wherein redundant self-confirmation statements follow correct reasoning steps, inflating output length and degrading inference efficiency. This paper formally defines the phenomenon and identifies a distinctive first-token probability bias as its statistical signature. Leveraging this insight, we propose a training-free, step-level compression method: reflection steps are localized via first-token statistics, and their outputs are suppressed without parameters. The method is fully compatible with vLLM and achieves 18.7% inference-length reduction out-of-the-box. When combined with lightweight fine-tuning, compression reaches 50.2%, with zero accuracy loss on R1-Distill, QwQ-32B, and Qwen3-32B. Our approach establishes a plug-and-play, theoretically grounded paradigm for efficient large-model inference.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Large Vision Models

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Large language models for search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

While recent advances in large reasoning models have demonstrated remarkable performance, efficient reasoning remains critical due to the rapid growth of output length. Existing optimization approaches highlights a tendency toward"overthinking", yet lack fine-grained analysis. In this work, we focus on Self-Affirmation Reflections: redundant reflective steps that affirm prior content and often occurs after the already correct reasoning steps. Observations of both original and optimized reasoning models reveal pervasive self-affirmation reflections. Notably, these reflections sometimes lead to longer outputs in optimized models than their original counterparts. Through detailed analysis, we uncover an intriguing pattern: compared to other reflections, the leading words (i.e., the first word of sentences) in self-affirmation reflections exhibit a distinct probability bias. Motivated by this insight, we can locate self-affirmation reflections and conduct a train-free experiment demonstrating that suppressing self-affirmation reflections reduces output length without degrading accuracy across multiple models (R1-Distill-Models, QwQ-32B, and Qwen3-32B). Furthermore, we also improve current train-based method by explicitly suppressing such reflections. In our experiments, we achieve length compression of 18.7% in train-free settings and 50.2% in train-based settings for R1-Distill-Qwen-1.5B. Moreover, our improvements are simple yet practical and can be directly applied to existing inference frameworks, such as vLLM. We believe that our findings will provide community insights for achieving more precise length compression and step-level efficient reasoning.

Problem

• Reduces output length by suppressing self-affirmation reflections
• Improves reasoning efficiency without accuracy degradation
• Enhances train-based methods for precise length compression

Innovation

• Suppress self-affirmation reflections in reasoning models
• Use leading word probability bias for detection
• Achieve length compression without accuracy loss