AlphaOne: Reasoning Models Thinking Slow and Fast at Test Time

Code

• github.com/astral-group/alphaone
  Official★ 88

Abstract

This paper presents AlphaOne (1), a universal framework for modulating reasoning progress in large reasoning models (LRMs) at test time. 1 first introduces moment, which represents the scaled thinking phase with a universal parameter . Within this scaled pre- moment phase, it dynamically schedules slow thinking transitions by modeling the insertion of reasoning transition tokens as a Bernoulli stochastic process. After the moment, 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 1's superior reasoning capability and efficiency. Project page: https://alphaone-project.github.io/

Tasks

Reproductions