Role Playing Leader Hallucination: A Lightweight Agent-based Multi-collaboration Framework

We hope to build an system that is lightweight, customizable, and that is caplable of social-reasoning and expecting what the other agent should be doing through the lens of advesarial collaboration. We are building a generalize agent model for every single agent in this environment and hopefully moving a step closer to level-one agent.

Image in curtesy of Zhiting Hu from UCSD HDSI: Level-1 Agent Concept (Cover page image in courtesy of Sichun Gao from UCSD CSE)

Check out our unpublished manuscript of RPLH: A Lightweight Agent-Based System For Multi-Collaboration Under Adversarial Settings.