Offline Reinforcement Learning on OGBench puzzle-3x3-singletask 5 tasks
43Normalized ScoreDeFlow
Evaluation Results
| Method | Links | |
|---|---|---|
| DeFlowPolicy Class=Multi-step / Iterative Policy2026.04 | 43 | |
| DROLPolicy Class=One-step Method, K=162026.04 | 41 | |
| DROLPolicy Class=One-step Method, K-selection protocol=benchmark-specific2026.04 | 41 | |
| FQLPolicy Class=One-step Method2026.04 | 30 | |
| ReBRACPolicy Class=Gaussian Policy2026.04 | 21 | |
| IFQLPolicy Class=Multi-step / Iterative Policy2026.04 | 19 | |
| FBRACPolicy Class=Multi-step / Iterative Policy2026.04 | 14 | |
| IQLPolicy Class=Gaussian Policy2026.04 | 9 | |
| FAWACPolicy Class=Multi-step / Iterative Policy2026.04 | 6 | |
| BCPolicy Class=Gaussian Policy2026.04 | 2 |