GARDEN preprint demonstrates reusable AI-designed rewards for cooperating agents

Researchers report in a preprint that GARDEN, a system that uses a language model to design rewards for cooperating agents, improved performance when its reward programs transferred unchanged to related scenarios unseen during design. The language model examines environment code, training records and agent behavior, then tests draft rewards in trial training runs. On two unseen Overcooked kitchen layouts, task returns increased by 58.2% and 52.5% over default rewards. In two SMAX combat scenarios, mean win rate rose from 51.0% to 89.9%, exceeding comparison methods under matched testing budgets. The demonstrations cover simulated kitchen and combat tasks, rather than deployed robots or autonomous vehicles.