The models of the paper "ProRAG: Process-Supervised Reinforcement Learning for Retrieval-Augmented Generation"