List your company

Offline Reinforcement Learning on D4RL hopper-medium-expert

119.2Normalized ScoreATAC
98.712104.031109.35114.669Jun 3, 2021Mar 30, 2022Jan 24, 2023Nov 20, 2023Sep 15, 2024Jul 12, 2025May 8, 2026

Evaluation results

MethodLinks
ATACSeeds=52023.10119.2—
ProposedSeeds=52023.10117.8—
NEUBAYMethod Category=Ours2025.12114.8—
VIPOMethod Category=Conservative model-based2025.12113.2—
PSPOAveraged across=4 random seeds2026.05112.8—
ADMPOMethod Category=Conservative model-based2025.12112.7—
AMGNumber of random seeds=42026.03112.7—
ADM2026.05112.7—
MOBILEMethod Category=Conservative model-based2025.12112.6—
RATE2023.06112.5—
TD3+BCSeeds=52023.10112.4—
CABI+TD3-BCMethod Type=Model-based2023.10112.4—
BC(CQL)Evaluation=Average score over 100 episodes2024.10111.9—
Fu et al.Algorithmic Template=Iterative2021.06111.9—
DMambasource=Lv et al., 20242023.06111.9—
RRPINumber of random seeds=42026.03111.9—
ATACMethod Type=Model-free2023.10111.9—
DDImplementation=Original2023.06111.8—
DDCategory=Diffusion Planner2024.05111.8—
Decision Diffuser2025.12111.8—
PMDBNumber of random seeds=42026.03111.8—
PMDBMethod Type=Model-based2023.10111.8—
PMDB2026.05111.8—
DDImplementation=reproduced2023.06111.6—
AdaptDiffuserCategory=Diffusion Planner2024.05111.6—
CBOPMethod Category=Bayesian-inspired2025.12111.6—
OptiDICESeeds=52023.10111.5—
CQLSeeds=52023.10111.4—
ROMI-CQL2021.10111.4—
Count-MORLMethod Type=Model-based2023.10111.4—
COMBOSeeds=52023.10111.1—
COMBO2021.10111.1—
COMBOCategory=Model-based RL2024.05111.1—
COMBOMethod Category=Model-free2025.12111.1—
COMBO2025.12111.1—
DMambasource=Ota, 20242023.06111.1—
DQLNFE=52026.03111.1—
COMBOMethod Type=Model-based2023.10111.1—
DMModel Category=SSM-based2026.02111—
BCQSeeds=52023.10110.9—
Decision Stacks2023.06110.9—
DS4Model Category=SSM-based2026.02110.8—
EDACMethod Category=Model-free2025.12110.7—
EDACMethod Type=Model-free2023.10110.7—
MambaDM2023.06110.5—
MPDiffuser+Rankranking=true2025.12110.4—
DMGNumber of random seeds=42026.03110.4—
DMG2026.05110.4—
DMMModel Category=SSM-based2026.02110.2—
DCModel Category=SSM-based2026.02110.1—
Trajectory Transformerdiscretization=quantile2021.06110—
TT2023.06110—
TT2023.06110—
TT2026.03110—
MPDiffuserranking=false2025.12109.9—
Planner2025.12109.5—
LEQMethod Category=Conservative model-based2025.12109.4—
MoMoMethod Category=Conservative model-based2025.12109.1—
CAMPAccess=Scripted preferences2024.04108.9—
EPQNumber of random seeds=42026.03108.8—
EPQ2026.05108.8—
MOREL2021.10108.7—
MOREL2023.06108.7—
MORELCategory=Model-based RL2024.05108.7—
DTModel Category=Transformer-based2026.02108.7—
MOReL2026.03108.7—
OPPOAccess=Scripted preferences, Implementation=Reported2024.04108—
SUMOMethod Category=Conservative model-based2025.12107.8—
Decision Transformer2021.06107.6—
DT2023.06107.6—
DT2025.12107.6—
DT2023.06107.6—
DT2026.03107.6—
CROPMethod Type=Model-based2023.10107.5—
IQLAccess=True reward functions2024.04107.4—
Diffuser2023.06107.2—
Diffuser2025.12107.2—
Decision Stacks2023.06107—
CQL2021.10106.8—
BAIL2021.10106.4—
Trajectory Transformerdiscretization=uniform2021.06106—
APE-VMethod Category=Conservative model-based2025.12105.7—
MOReLNumber of random seeds=42026.03105.6—
MOReL2026.05105.6—
TAP2023.06105.5—
CQL2021.06105.4—
CQL2023.06105.4—
CQLModel Category=Value-based2026.02105.4—
CQL2023.06105.4—
CQL2026.03105.4—
IDQL2025.12105.3—
DT2023.06105.2—
EDTModel Category=Transformer-based2026.02104.7—
KDPNFE=12026.03103.6—
DiffuserCategory=Diffusion Planner2024.05103.3—
DiffuserNFE=202026.03103.3—
MROracle status=false, Implementation source=reimpl.2023.05103—
Rev. KL Reg.Algorithmic Template=One-step2021.06102.1—
CACNFE=22026.03100.4—
OPPOAccess=Scripted preferences, Implementation=Re-implementation2024.0499.5—