Offline Reinforcement Learning Efficiency on D4RL Antmaze v0
0.01Runtime (s)DiffuserLite-R2
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DiffuserLite-R2Backbone=rectified flow with reflow step2024.01 | 0.01 | 101.7 | |
| DiffuserLite-R1Backbone=rectified flow2024.01 | 0.015 | 65.7 | |
| DiffuserLite-DBackbone=diffusion model2024.01 | 0.027 | 37.3 | |
| HDMI2024.01 | 0.41 | 2.4 | |
| Diffuser2024.01 | 0.791 | 1.3 | |
| DD2024.01 | 2.591 | 0.39 |