Multi-agent coordination on RedBlueDoors (test)
0.984Average Rewardae-comm
Evaluation results
| Method | Links | |
|---|---|---|
| ae-commtraining_method=autoencoded communication2021.10 | 0.984 | |
| rl-comm w/ biastraining_method=reinforcement learning, inductive_biases=positive signaling and listening2021.10 | 0.729 | |
| rl-commtraining_method=reinforcement learning2021.10 | 0.174 | |
| no-commcommunication=none2021.10 | 0.123 |