Paper 2602.08354
Does Your Reasoning Model Implicitly Know When to Stop Thinking?
- Published
- Feb 2026
- Research lab
- Independent
- Citations
- 20
- GitHub
- Not linked
01 In brief
Summary
The paper investigates why large reasoning models (LRMs) produce lengthy, redundant chains of thought (CoTs) despite shorter chains often being more accurate.
The authors discover that LRMs implicitly know when to stop thinking, but this capability is obscured by current sampling paradigms like pass@1.
They introduce SAGE (Self-Aware Guided Efficient Reasoning), a sampling strategy that uses cumulative log-probability to explore reasoning steps and select high-confidence, concise chains.
Integrating SAGE into reinforcement learning (SAGE-RL) enables models to learn these efficient patterns, improving both accuracy and token efficiency across benchmarks like MATH-500, AIME 2024/2025, and OlympiadBench.
Experiments show SAGE-RL reduces response length by up to 44.1% while improving accuracy by an average of 2.1% compared to baselines, with minimal training overhead.
The work demonstrates that LRMs possess an inherent sense of when to stop reasoning, which can be leveraged for more efficient inference without sacrificing performance.
02 From the paper
Abstract
Recent advancements in large reasoning models (LRMs) have greatly improved their capabilities on complex reasoning tasks through Long Chains of Thought (CoTs). However, this approach often results in substantial redundancy, impairing computational efficiency and causing significant delays in real-time applications. Recent studies show that longer reasoning chains are frequently uncorrelated with correctness and can even be detrimental to accuracy. In a further in-depth analysis of this phenomenon, we surprisingly uncover and empirically verify that LRMs implicitly know the appropriate time to stop thinking, while this capability is obscured by current sampling paradigms. Motivated by this, we introduce SAGE (Self-Aware Guided Efficient Reasoning), a novel sampling paradigm that unleashes this efficient reasoning potential. Furthermore, integrating SAGE as mixed sampling into group-based reinforcement learning (SAGE-RL) enables SAGE-RL to effectively incorporate SAGE-discovered efficient reasoning patterns into standard pass@1 inference, markedly enhancing both the reasoning accuracy and efficiency of LRMs across multiple challenging mathematical benchmarks.