Offline Reinforcement Learning on OGBench cube-double-singletask 5 tasks
40Normalized ScoreDeFlow
Evaluation results
| Method | Links | |
|---|---|---|
| DeFlowPolicy Class=Multi-step / Iterative Policy2026.04 | 40 | |
| FQLPolicy Class=One-step Method2026.04 | 29 | |
| DROLPolicy Class=One-step Method, K-selection protocol=benchmark-specific2026.04 | 26 | |
| SORLPolicy Class=Multi-step / Iterative Policy2026.04 | 25 | |
| DROLPolicy Class=One-step Method, K=162026.04 | 20 | |
| FBRACPolicy Class=Multi-step / Iterative Policy2026.04 | 15 | |
| IFQLPolicy Class=Multi-step / Iterative Policy2026.04 | 14 | |
| ReBRACPolicy Class=Gaussian Policy2026.04 | 12 | |
| IQLPolicy Class=Gaussian Policy2026.04 | 7 | |
| FAWACPolicy Class=Multi-step / Iterative Policy2026.04 | 5 | |
| BCPolicy Class=Gaussian Policy2026.04 | 2 |