Single-agent Reinforcement Learning (SARL) on Pong Poisoned
18.2Average Episode ReturnBehaviorGuard
Evaluation Results
| Method | Links | |
|---|---|---|
| BehaviorGuard2026.05 | 18.2 | |
| SHINE2026.05 | 17.9 | |
| STRIP2026.05 | 15.6 | |
| FeatureRE2026.05 | 4.9 | |
| Direct retraining2026.05 | 3 | |
| Original2026.05 | 2.1 | |
| NC2026.05 | 0.1 |