One Missing Piece for Open-Source Reasoning Models: A Dataset to Mitigate Cold-Starting Short CoT LLMs in RL

TL;DR

This paper introduces the Long CoT Collection, a large dataset of 100K reasoning rationales designed to improve independent reasoning capabilities in large language models, facilitating better training and reinforcement learning performance.

Contribution

The paper presents a novel dataset and pipeline for inducing reasoning strategies into short CoT LLMs, enabling longer, controllable reasoning and enhancing RL training outcomes.

Findings

Dataset quality is comparable to existing models.

Training on the dataset improves reasoning skills.

Models initialized on the dataset show 2-3x larger RL gains.

Abstract

With the release of R1, a publicly available large reasoning model (LRM), researchers commonly train new LRMs by training language models on R1's long chain-of-thought (CoT) inferences. While prior works show that LRMs' capabilities can be reproduced through direct distillation, the continued reliance on the existing models (e.g., R1) remains a critical limitation in advancing the field. As a first step toward independent LRM development, this paper explores the possibility of constructing a long CoT dataset with LLMs that are not trained for inference-time scaling. To this end, we present the Long CoT Collection, a dataset of 100K CoT rationales annotated using existing short CoT LLMs. We develop a pipeline that induces o1's novel reasoning strategies into short CoT LLMs, enabling them to think longer and introducing controllability over the thought budget to better manage the overthinking problem. Our extensive analyses validate that our dataset achieves quality comparable to--or slightly below--R1. Furthermore, our experiments demonstrate that training on our dataset not only strengthens general reasoning skills, but also provides a strong foundation for reinforcement learning--models initialized on our data achieve 2-3x larger gains with RLVR.