open-thought/reasoning-gym
A Python library of RL environments with algorithmically verifiable rewards for training reasoning capabilities in language models.
Not currently ranked — collecting fresh signals.
star history
Reasoning Gym provides procedurally generated dataset generators and verifiable reasoning environments for training and evaluating reasoning models using reinforcement learning. It offers over 100 tasks across diverse domains including algebra, geometry, graph theory, logic, and games. Tasks are designed with single or multiple correct solutions, with a standard interface for procedural verification of model outputs.
Frequently asked
- What is open-thought/reasoning-gym?
- A Python library of RL environments with algorithmically verifiable rewards for training reasoning capabilities in language models.
- Is reasoning-gym open source?
- Yes — open-thought/reasoning-gym is open source, released under the Apache-2.0 license.
- What language is reasoning-gym written in?
- open-thought/reasoning-gym is primarily written in Python.
- How popular is reasoning-gym?
- open-thought/reasoning-gym has 1.5k stars on GitHub.
- Where can I find reasoning-gym?
- open-thought/reasoning-gym is on GitHub at https://github.com/open-thought/reasoning-gym.