List your company

Offline Reinforcement Learning on D4RL (Full Suite Evaluation)

65.9HalfCheetah-MediumEDAC
-0.76416.54333.8551.157Jun 9, 2023Oct 22, 2023Mar 5, 2024Jul 18, 2024Nov 30, 2024Apr 14, 2025Aug 27, 2025

Evaluation results

MethodLinks
EDACRuntime (s/epoch)=19.6, GPU Memory (GB)=1.92023.0665.928.425.316.6101.692.561.310187.1106.3110.7114.7106.8110.1115.11,243.416.70000016.752.168.2120.31,380.4—————————————————————
CPI-RERuntime (s/epoch)=19.1, GPU Memory (GB)=1.42023.0665.930.730.45.597.986.355.9103.293.895.6110.1111.297.4112.3111.21,207.499.292.684.880.633.648438.88770.7157.71,803.9—————————————————————
ReBRACHyperparameter configuration=dataset-specific, Robust Critic=false2025.0865.629.58.118.410282.55198.177.3101.1107111.6105.9100.1112.3—106.888.38476.360.454.4—103.591.8——82.1107.1102.281.2105.178.396.776.8154.101.1104.60.26.7133.800.9106.658.674.8—
CPIRuntime (s/epoch)=8.5, GPU Memory (GB)=1.42023.0664.429.729.55.998.585.854.6101.791.894.7106.4110.996.5112.2110.61,193.298.888.682.480.420.645.241680.171.8151.91,761.2—————————————————————
ASPCHyperparameter configuration=fixed, Robust Critic=false2025.0858.720.89.415.692.792.450.6101.397.699.9111111.1105.1112.3110.8—128.192.585705552.5—81.187.2——79.3107.2102.182.1117.8195.8147.274.5141.200105.62.212128.60.10111.255.777.9—
A2PRHyperparameter configuration=fixed, Robust Critic=false2025.0856.121.120.11.278.384.24998.984.595.9110.888.299.983.984.8—102.532.54040522.5—-2.16.5——79.597.1102.574.290.4177.7123.5338.7551.5-0.2-0.3-0.31.10.30.3-0.3-0.3-0.34.6551.2—
wPCHyperparameter configuration=fixed, Robust Critic=true2025.0854.818.88.512.581.889.648.1100.893.498.9109.1110.1103.879.1111.5—73.17585856565—39.934.6——76.7105.699.577.887.4123.394.678.7141.8-0.2051.400.157.60.10.16.728.864.2—
Diffusion-QL (D-QL)Runtime (s/epoch)=39.8, GPU Memory (GB)=1.52023.0651.52218.35.596.687.348.31029897.2112.3111.296.3102.6109.51,158.6968479.8824961.7452.575.760.8136.51,747.6—————————————————————
IQLHyperparameter configuration=dataset-specific, Robust Critic=false2025.085019.510.111.365.280.742.189.675.492.785.5112.195.5108.896.9—-8.970.664.661.742.527.6—81.577.2——75104.497.572.934.861.746.258.3133.63.10.8105.32.51.1129.60.10.2106.553.462.6—
TD3+BCHyperparameter configuration=fixed, Robust Critic=true2025.0849.610.68.65.9626245.847.476.697.9103.3103.3100.4108.2108.2—34.587.57.512.52.52.5—53.871.7——74.590.388.370.763.3108.968.935.4126.60081.600.1132.80090.646.457.7—
POR2023.0648.829126.398.281.143.598.976.694.790109.193.2110.4102.91,094.776.864.887.275.224.459.2387.676.967.6144.51,626.8—————————————————————
TD3+BCRuntime (s/epoch)=7.4, GPU Memory (GB)=1.42023.0648.3118.51.659.383.744.660.981.890.798110.196.7107.8110.21,013.278.671.4310.600.2163.8-1.99.67.71,184.7—————————————————————
InAC2023.0648.319.632.46.360.382.744.392.169.883.593.810993.6103.4110.61,049.784.882.4—————52.3-844.3——————————————————————
Cal-QL (OffRL Whole)Training Data=Whole offline dataset (D_off)2024.1047.6———56.48445.89887.450.8107.3109———70.1————————————————————————————————
IQL2023.0647.413.78.45.966.378.344.294.773.986.791.5109.694.9108.8109.71,03487.562.271.27039.647.537871.537.3108.81,520.8—————————————————————
CQL2023.0646.931.35.35.461.979.545.386.376.89596.9109.197.3106.5109.31,052.8748461.253.715.814.9303.635.227.262.41,418.8—————————————————————
DT2023.0642.62.25.42.267.67436.682.766.686.8107.6108.187.794.2108.3972.659.2530000112.273.967.3141.21,226—————————————————————
BC (Whole)Training Data=Whole offline dataset (D_off)2024.1042.3———50.265.91422.914.659.851.789.6———45.7————————————————————————————————
Av-PBCTraining Data=Synthetic (N_syn=256)2024.1036.9———32.539.535.940.9552238.742.1———38.2————————————————————————————————
PBCTraining Data=Synthetic (N_syn=256)2024.1030.9———25.133.218.435.641.520.533.434———30.4————————————————————————————————
DBCTraining Data=Synthetic (N_syn=256)2024.1028.2———37.829.32.512.16.12931.111.7———20.9————————————————————————————————
RandomTraining Data=Synthetic (N_syn=256), Selection=Randomly selecting real state-action pairs2024.101.8———19.24.90.919.11.9211.66.7———7.6————————————————————————————————
BCPolicy Category=Gaussian2026.05———————————————————————————————————————————————38
DriftQLPolicy Category=Drift2026.05———————————————————————————————————————————————72
FQLAlgorithm=FQL2025.12——————————————————————84——5262.1—————————————————————
FQLPolicy Category=Flow2026.05———————————————————————————————————————————————66
GFP actorAlgorithm=GFP, Policy Component=actor πθ2025.12——————————————————————83.1——52.863—————————————————————
GFP VaBCAlgorithm=GFP, Policy Component=VaBC πω2025.12——————————————————————70.2——49.656.5—————————————————————
IDQLPolicy Category=Diffusion2026.05———————————————————————————————————————————————71
IFQLPolicy Category=Flow2026.05———————————————————————————————————————————————56
IQLAlgorithm=IQL2025.12——————————————————————17——4854—————————————————————
IQLPolicy Category=Gaussian2026.05———————————————————————————————————————————————64
ReBRACAlgorithm=ReBRAC2025.12——————————————————————76.8——5964.8—————————————————————
ReBRACPolicy Category=Gaussian2026.05———————————————————————————————————————————————76
SRPOPolicy Category=Diffusion2026.05———————————————————————————————————————————————71