The Berkeley RL Club That Gives Every Major a Dataset

Field notes

The club divides its work between building reinforcement-learning environments and supplying expert judgments that models cannot reliably generate on their own. Its examples span argument evaluation, ethics rubrics, robotics, and language-model training, with students grouped by the kind of expertise they can contribute. The homepage also describes a reading group, research pods, an applied RL DeCal, and an engineering track intended to deliver data in formats AI labs already use.