Open-source projects tagged “reinforcement-learning”

Open-source toolkit from Unity that lets games and simulations serve as environments for training AI agents with reinforcement and imitation learning.

A short introductory book on reinforcement learning, with PyTorch implementations of the algorithms it covers from MC to PPO.

Converts any Agent Skill into RL-ready terminal tasks in Harbor format, using Codex agents to plan and generate them.

ASPIRE is a research framework where coding agents iteratively write and refine sensorimotor skills for robot manipulation tasks in simulation and on real robots.

A domain-general framework that uses world-model predictions and interaction feedback for recursive self-improvement across code repair, GUI tasks and lab planning.