Revisiting Reinforcement Learning for LLM Reasoning from A Cross-Domain Perspective

TL;DR

This paper introduces Guru, a large, diverse RL reasoning corpus, and demonstrates that domain-specific RL training enhances reasoning skills in large language models, especially in less-pretrained domains, leading to state-of-the-art results.

Contribution

The creation of Guru, a comprehensive RL reasoning dataset across six domains, and the systematic analysis of RL's effectiveness in improving LLM reasoning in various domains.

Findings

RL benefits domain-rich pretraining areas like Math and Code.

In-domain RL training is crucial for less-pretrained domains like Logic and Simulation.

Proposed models outperform existing baselines on a multi-domain reasoning benchmark.

Abstract

Reinforcement learning (RL) has emerged as a promising approach to improve large language model (LLM) reasoning, yet most open efforts focus narrowly on math and code, limiting our understanding of its broader applicability to general reasoning. A key challenge lies in the lack of reliable, scalable RL reward signals across diverse reasoning domains. We introduce Guru, a curated RL reasoning corpus of 92K verifiable examples spanning six reasoning domains--Math, Code, Science, Logic, Simulation, and Tabular--each built through domain-specific reward design, deduplication, and filtering to ensure reliability and effectiveness for RL training. Based on Guru, we systematically revisit established findings in RL for LLM reasoning and observe significant variation across domains. For example, while prior work suggests that RL primarily elicits existing knowledge from pretrained models, our results reveal a more nuanced pattern: domains frequently seen during pretraining (Math, Code, Science) easily benefit from cross-domain RL training, while domains with limited pretraining exposure (Logic, Simulation, and Tabular) require in-domain training to achieve meaningful performance gains, suggesting that RL is likely to facilitate genuine skill acquisition. Finally, we present Guru-7B and Guru-32B, two models that achieve state-of-the-art performance among open models RL-trained with publicly available data, outperforming best baselines by 7.9% and 6.7% on our 17-task evaluation suite across six reasoning domains. We also show that our models effectively improve the Pass@k performance of their base models, particularly on complex tasks less likely to appear in pretraining data. We release data, models, training and evaluation code to facilitate general-purpose reasoning at: https://github.com/LLM360/Reasoning360