List your company
ResearchFormerly WizWand

Continuous Control on Pendulum (Robustness Gap)

0.72Robustness GapUP(lstm)
0.68920.89711.1051.3129Apr 30, 2026
Updated 6mo ago

Evaluation results

MethodLinks
UP(lstm)Context=g, Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.040.72
UP(gru)Context=g, Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.040.72
UP(mlp(wide))Context=g, Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.040.8
UP(lstm)Context=l, Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.040.84
UP(gru)Context=l, Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.040.85
UP(mlp(narrow))Context=l, Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.040.87
UP(mlp(narrow))Context=g, Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.040.95
UP(mlp(wide))Context=l, Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.041.01
UP(gru)Context=(g,l), Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.041.11
UP(lstm)Context=(g,l), Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.041.24
UP(mlp(wide))Context=(g,l), Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.041.3
UP(mlp(narrow))Context=(g,l), Training Algorithm=DDPG, State-action history length (k)=4, Data retention fraction (alpha)=Tuned2026.041.49