Multi-agent pathfinding on POGEMA
2.16Average Throughput (Random)Expert
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ExpertTask Specificity=Single-task expert2026.04 | 2.16 | 1.55 | — | — | |
| MARL-GPTTraining Scope=Multi-Env, Memory Architecture=Memory-based, Number of Parameters=7M2026.04 | 1.16 | 0.96 | 1.02 | 2.72 | |
| CQLTraining Scope=Single-Env, Memory Architecture=Memory-free2026.04 | 0.26 | 0.14 | 0.17 | 0.66 | |
| BCTraining Scope=Single-Env, Memory Architecture=Memory-free2026.04 | 0.24 | 0.1 | 0.14 | 0.46 | |
| BC-LSTMTraining Scope=Single-Env, Memory Architecture=Memory-based2026.04 | 0.23 | 0.11 | 0.12 | 0.49 | |
| DTTraining Scope=Single-Env, Memory Architecture=Memory-based2026.04 | 0.22 | 0.12 | 0.13 | 0.68 | |
| RATETraining Scope=Single-Env, Memory Architecture=Memory-based2026.04 | 0.21 | 0.12 | 0.13 | 0.65 |