Offline-to-online Reinforcement Learning on Robomimic and OGBench
95.8Average Success RateDFP
Evaluation Results
| Method | Links | |
|---|---|---|
| DFPinference_type=teacher-free one-step, action_chunking=true2026.05 | 95.8 | |
| QC-FQLinference_type=distilled one-step, action_chunking=true2026.05 | 86.8 | |
| QC-BFNinference_type=multi-step, action_chunking=true2026.05 | 83.8 | |
| MVPinference_type=teacher-free one-step, action_chunking=true2026.05 | 80.3 | |
| BFNinference_type=multi-step, action_chunking=false2026.05 | 38.4 | |
| FQLinference_type=distilled one-step, action_chunking=false2026.05 | 30.3 |