Cooperative Multi-agent Reinforcement Learning (coo-MARL) on COMA Poisoned
94.2Average Winning RateBehaviorGuard
Evaluation results
| Method | Links | |
|---|---|---|
| BehaviorGuard2026.05 | 94.2 | |
| BIRD2026.05 | 93.6 | |
| SHINE2026.05 | 88.1 | |
| Direct retraining2026.05 | 82.3 | |
| STRIP2026.05 | 71.3 | |
| Original2026.05 | 35.7 | |
| FeatureRE2026.05 | 24.7 | |
| NC2026.05 | 21.6 |