Google OpenRL makes LLM training more controllable
… be able to write training logic without rebuilding all infrastructure for samplers, trainers, jobs and Kubernetes each time. This is not a consumer feature and not a chatbot announcement. It is infrastructure for teams …