REASONING GYM: Reasoning Environments for Reinforcement Learning with Verifiable Rewards
This addresses the need for scalable and verifiable environments for reinforcement learning in reasoning, though it is incremental as it builds on existing procedural generation approaches.
The authors tackled the problem of limited training data for reinforcement learning in reasoning tasks by introducing Reasoning Gym, a library that provides over 100 data generators and verifiers across multiple domains, enabling virtually infinite training data with adjustable complexity. Their experimental results show its efficacy in evaluating and training reasoning models.
We introduce Reasoning Gym (RG), a library of reasoning environments for reinforcement learning with verifiable rewards. It provides over 100 data generators and verifiers spanning multiple domains including algebra, arithmetic, computation, cognition, geometry, graph theory, logic, and various common games. Its key innovation is the ability to generate virtually infinite training data with adjustable complexity, unlike most previous reasoning datasets, which are typically fixed. This procedural generation approach allows for continuous evaluation across varying difficulty levels. Our experimental results demonstrate the efficacy of RG in both evaluating and reinforcement learning of reasoning models.