Offline Reinforcement Learning on OGBench humanoidmaze-medium-singletask (5 tasks)
68Normalized ScoreDROL
Evaluation Results
| Method | Links | |
|---|---|---|
| DROLPolicy Class=One-step Method, K-selection protocol=benchmark-specific2026.04 | 68 | |
| DROLPolicy Class=One-step Method, K=162026.04 | 65 | |
| SORLPolicy Class=Multi-step / Iterative Policy2026.04 | 64 | |
| IFQLPolicy Class=Multi-step / Iterative Policy2026.04 | 60 | |
| FQLPolicy Class=One-step Method2026.04 | 58 | |
| DeFlowPolicy Class=Multi-step / Iterative Policy2026.04 | 48 | |
| FBRACPolicy Class=Multi-step / Iterative Policy2026.04 | 38 | |
| IQLPolicy Class=Gaussian Policy2026.04 | 33 | |
| ReBRACPolicy Class=Gaussian Policy2026.04 | 22 | |
| FAWACPolicy Class=Multi-step / Iterative Policy2026.04 | 19 | |
| BCPolicy Class=Gaussian Policy2026.04 | 2 |