Tag
A paper proposing a mechanism for enterprise agents to improve by safely converting messy daily work into learning data, using a data proxy and control layer, with AREAL2.0 demonstrating online RL from real interaction traces.