Controlling Thinking Speed in Reasoning Models

TL;DR

This paper introduces a method for Large Reasoning Models to dynamically adjust their thinking speed, balancing accuracy and efficiency by controlling internal representations and estimating problem difficulty in real-time.

Contribution

It presents the first approach to control reasoning speed in LRMs through representation editing and real-time difficulty estimation, improving performance without additional training.

Findings

Achieves +1.3% accuracy on reasoning benchmarks

Reduces token usage by 8.6% on average

Outperforms existing prompt-based scaling methods

Abstract

Human cognition is theorized to operate in two modes: fast, intuitive System 1 thinking and slow, deliberate System 2 thinking. While current Large Reasoning Models (LRMs) excel at System 2 thinking, their inability to perform fast thinking leads to high computational overhead and latency. In this work, we enable LRMs to approximate human intelligence through dynamic thinking speed adjustment, optimizing accuracy-efficiency trade-offs. Our approach addresses two key questions: (1) how to control thinking speed in LRMs, and (2) when to adjust it for optimal performance. For the first question, we identify the steering vector that governs slow-fast thinking transitions in LRMs' representation space. Using this vector, we achieve the first representation editing-based test-time scaling effect, outperforming existing prompt-based scaling methods. For the second question, we apply real-time difficulty estimation to signal reasoning segments of varying complexity. Combining these techniques, we propose the first reasoning strategy that enables fast processing of easy steps and deeper analysis for complex reasoning. Without any training or additional cost, our plug-in module delivers an average +1.3% accuracy with -8.6% token usage across leading LRMs and advanced reasoning benchmarks. All of our algorithms are implemented based on vLLM and are expected to support broader applications and inspire future research.