S3-CoT: Self-Sampled Succinct Reasoning Enables Efficient Chain-of-Thought LLMs

TL;DR

This paper introduces S3-CoT, a self-sampling framework that enables efficient chain-of-thought reasoning in large language models by reducing reliance on high-quality supervision data and mimicking human-like fast thinking.

Contribution

The paper proposes a novel self-sampling method for training LLMs to perform efficient CoT reasoning without extensive labeled data, using activation steering and a progressive curriculum.

Findings

Improved performance on math benchmarks and cross-domain tests.

Stable enhancements for both general and R1-style LLMs.

Effective reasoning trace induction without teacher guidance.

Abstract

Large language models (LLMs) equipped with chain-of-thought (CoT) achieve strong performance and offer a window into LLM behavior. However, recent evidence suggests that improvements in CoT capabilities often come with redundant reasoning processes, motivating a key question: Can LLMs acquire a fast-thinking mode analogous to human System 1 reasoning? To explore this, our study presents a self-sampling framework based on activation steering for efficient CoT learning. Our method can induce style-aligned and variable-length reasoning traces from target LLMs themselves without any teacher guidance, thereby alleviating a central bottleneck of SFT-based methods-the scarcity of high-quality supervision data. Using filtered data by gold answers, we perform SFT for efficient CoT learning with (i) a human-like dual-cognitive system, and (ii) a progressive compression curriculum. Furthermore, we explore a self-evolution regime in which SFT is driven solely by prediction-consistent data of variable-length variants, eliminating the need for gold answers. Extensive experiments on math benchmarks, together with cross-domain generalization tests in medicine, show that our method yields stable improvements for both general and R1-style LLMs. Our data and model checkpoints can be found at https://github.com/DYR1/S3-CoT.