Offline Reinforcement Learning for Plasma Control in Nuclear Fusion: Codebase and Benchmark

In three linesRL4F is an open-source offline reinforcement learning benchmark for plasma control in nuclear fusion. Built on historical data from the DIII-D tokamak, it evaluates imitation learning and offline RL methods on four multi-actuator tracking tasks (rotation, density, temperature, pressure). Offline model-based RL methods achieve best average performance.

Summary generated by Claude — human-verified