RESEARCH · RESEARCH · #2876
GARDEN: LLM-driven generalist reward design for cooperative multi-agent RL
The paper introduces GARDEN (Generalist Autonomous Reward Design through Evidence Navigation), an approach using large language models as autonomous reward engineers to produce a single reward program that transfers unchanged across related cooperative multi-agent reinforcement learning scenarios. On unseen Overcooked and SMAX scenarios, GARDEN increased task returns by 58.2% and 52.5% over default rewards and raised mean win rate in two combat scenarios from 51.0% to 89.9%, outperforming fixed diagnostic feedback and Eureka-inspired reward reflection under matched budgets.
KEY POINTS
- The paper introduces GARDEN (Generalist Autonomous Reward Design through Evidence Navigation), an approach using large language models as autonomous reward engineers to produce a single reward program that transfers unchanged across related cooperative multi-agent reinforcement learning scenarios.
- On unseen Overcooked and SMAX scenarios, GARDEN increased task returns by 58.2% and 52.5% over default rewards and raised mean win rate in two combat scenarios from 51.0% to 89.9%, outperforming fixed diagnostic feedback and Eureka-inspired reward reflection under matched budgets.
- Demonstrates that LLMs can autonomously engineer reusable reward programs that transfer across related cooperative MARL tasks, reducing manual reward design and materially improving performance.
WHY IT MATTERS
Demonstrates that LLMs can autonomously engineer reusable reward programs that transfer across related cooperative MARL tasks, reducing manual reward design and materially improving performance.