SATURN: SAT-based Reinforcement Learning to Unleash LLMs Reasoning

How to design reinforcement learning (RL) tasks that effectively unleash the reasoning capability of large language models (LLMs) remains an open question. Existing RL tasks (e.g., math, programming, and constructing reasoning tasks) face three key limitations: ❶ Scalability. They rely heavily on human annotation or expensive LLM synthesis to generate sufficient training data. ❷ Verifiability. LLMs' outputs are hard to verify automatically and reliably. ❸ Controllable Difficulty. Most tasks lack fine-grained difficulty control, making it challenging to train LLMs from easy to hard and progressively develop reasoning capability. To address these limitations, we propose SATURN, a SAT-based RL framework that uses Boolean Satisfiability (SAT) problems to train and evaluate LLM reasoning. SATURN enables scalable task construction, rule-based verification, and precise difficulty control. SATURN designs a curriculum learning pipeline that continuously improves LLMs' reasoning capability by constructing SAT tasks of increasing difficulty and training LLMs from easy to hard. To ensure stable training, we design a principled mechanism to control difficulty transitions. We introduce SATURN-2.6k, a dataset of 2,660 SAT problems with varying difficulty. It supports the evaluation of how LLM reasoning changes with problem difficulty. We apply SATURN to DeepSeek-R1-Distill-Qwen and obtain SATURN-1.5B and SATURN-7B. We achieve several notable results: ❶ On SAT problems, SATURN-1.5B and SATURN-7B achieve average pass@3 improvements of +14.0 and +28.1, respectively. ❷ On math and programming tasks, SATURN-1.5B and SATURN-7B improve average scores by +4.9 and +1.8 on benchmarks (e.g., AIME, LiveCodeBench). ❸ Compared to the state-of-the-art (SOTA) approach in constructing RL tasks, SATURN achieves further improvements of +8.8%. We release the source code, data, and models to support future research at https: //github.com/gtxygyzb/Saturn-code.