GenReasoning Debuts RLCR Framework to Build Collaborative AI Agent Organizations

GenReasoning introduced the Reinforcement Learning from Collective Rewards (RLCR) framework, which trains language models into collaborative agent organizations sharing a workspace and message board. Using the method to train the 120B parameter Avalon-M1-Mini, the company reported the emergence of organizational hierarchies, divisions of labor, and specialized testers and verifiers, with the framework intended as a laboratory for studying collective intelligence and multi-agent system safety.