List your company

Single-agent Reinforcement Learning (SARL) on Breakout Clean

505.5Average Episode ReturnSHINE
200.468279.659358.85438.041May 7, 2026

Evaluation results

MethodLinks
SHINE2026.05505.5
BIRD2026.05478.2
STRIP2026.05476.8
BehaviorGuard2026.05465.2
FeatureRE2026.05450.1
Original2026.05445.3
Direct retraining2026.05435.1
NC2026.05252.5
PD2026.05212.2