Communication-based Multi-agent Reinforcement Learning (com-MARL) on YSNP Poisoned
47.8Average Winning RateBehaviorGuard
Evaluation results
| Method | Links | |
|---|---|---|
| BehaviorGuard2026.05 | 47.8 | |
| SHINE2026.05 | 47.2 | |
| BIRD2026.05 | 43.6 | |
| Direct retraining2026.05 | 27.7 | |
| Original2026.05 | 18.3 |