Single-agent Reinforcement Learning (SARL) on Pong Clean
18.5Average Episode ReturnBehaviorGuard
Evaluation Results
| Method | Links | |
|---|---|---|
| BehaviorGuard2026.05 | 18.5 | |
| SHINE2026.05 | 18.2 | |
| Original2026.05 | 16.9 | |
| STRIP2026.05 | 15.2 | |
| FeatureRE2026.05 | 7.3 | |
| Direct retraining2026.05 | 7.1 | |
| NC2026.05 | 3.4 |