Scalable Evaluation of Multi-Agent Reinforcement Learning with Melting Pot
This addresses the problem of evaluating generalization in MARL for researchers, though it is incremental as it builds on existing evaluation methods by focusing on novel scenarios.
The paper tackles the lack of evaluation suites for multi-agent reinforcement learning (MARL) that assess generalization to novel situations, introducing Melting Pot, which includes over 80 unique test scenarios and reveals weaknesses in standard MARL algorithms not apparent from training performance.
Existing evaluation suites for multi-agent reinforcement learning (MARL) do not assess generalization to novel situations as their primary objective (unlike supervised-learning benchmarks). Our contribution, Melting Pot, is a MARL evaluation suite that fills this gap, and uses reinforcement learning to reduce the human labor required to create novel test scenarios. This works because one agent's behavior constitutes (part of) another agent's environment. To demonstrate scalability, we have created over 80 unique test scenarios covering a broad range of research topics such as social dilemmas, reciprocity, resource sharing, and task partitioning. We apply these test scenarios to standard MARL training algorithms, and demonstrate how Melting Pot reveals weaknesses not apparent from training performance alone.