ResearchBenchmarksGoal-conditioned navigation on MiniGrid 8x8 RGB image input (test)Follow0.43Clean RewardTDRT-PPO0.36760.38380.40.4162Jun 6, 2024Evaluation ResultsMethodMethodLinksClean RewardAttack Reward (SA-RL)Attack Reward (PA-AD)Attack Reward (BIA-ILfD, ours)Attack Reward (Avg)TDRT-PPOTarget Reward=0.96 ± 0Target Reward=0.96 ± 02024.060.430.320.580.290.4PPOTarget Reward=0.96 ± 0...Target Reward=0.96 ± 0, Defense Status=No defense2024.060.370.50.920.480.63