AlphaOne: Reasoning Models Thinking Slow and Fast at Test Time

TL;DR

AlphaOne introduces a universal framework that dynamically balances slow and fast reasoning in large models at test time, improving reasoning accuracy and efficiency across diverse tasks.

Contribution

It proposes a novel scalable reasoning modulation method using the alpha moment and stochastic scheduling, unifying and extending existing approaches.

Findings

Outperforms existing methods on mathematical benchmarks

Enhances reasoning efficiency without sacrificing accuracy

Demonstrates versatility across multiple scientific domains

Abstract

This paper presents AlphaOne ($\alpha$1), a universal framework for modulating reasoning progress in large reasoning models (LRMs) at test time. $\alpha$1 first introduces $\alpha$ moment, which represents the scaled thinking phase with a universal parameter $\alpha$. Within this scaled pre-$\alpha$ moment phase, it dynamically schedules slow thinking transitions by modeling the insertion of reasoning transition tokens as a Bernoulli stochastic process. After the $\alpha$ moment, $\alpha$1 deterministically terminates slow thinking with the end-of-thinking token, thereby fostering fast reasoning and efficient answer generation. This approach unifies and generalizes existing monotonic scaling methods by enabling flexible and dense slow-to-fast reasoning modulation. Extensive empirical studies on various challenging benchmarks across mathematical, coding, and scientific domains demonstrate $\alpha$1's superior reasoning capability and efficiency. Project page: https://alphaone-project.github.io/