Agents receive detailed feedback on their performance after each round, which not only influences rewards but guides future improvements. This mechanism functions like a self-improving loop, akin to reinforcement learning from competition performance. @FractionAI