Offline Reinforcement Learning on D4RL Franka Kitchen
85Mixed Success RateTRA
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| TRARepresentation Strategy=TRA, MVL=false2026.02 | 85 | 43 | 72 | — | — | — | |
| VIB-MVLRepresentation Strategy=VIB, MVL=true2026.02 | 84 | 52 | 85 | — | — | — | |
| VIBRepresentation Strategy=VIB, MVL=false2026.02 | 81 | 56 | 75 | — | — | — | |
| ORIG-MVLRepresentation Strategy=Original (GCIVL), MVL=true2026.02 | 79 | 59 | 81 | — | — | — | |
| ORIGRepresentation Strategy=Original (GCIVL), MVL=false2026.02 | 77 | 54 | 81 | — | — | — | |
| DUALRepresentation Strategy=DUAL, MVL=false2026.02 | 77 | 46 | 81 | — | — | — | |
| DUAL-MVLRepresentation Strategy=DUAL, MVL=true2026.02 | 76 | 46 | 78 | — | — | — | |
| DD2025.10 | 75 | — | 56.5 | 65.8 | — | — | |
| MDD2025.10 | 75 | — | 99.7 | 87.4 | — | — | |
| DiffuserLiteBackbone=D, Runtime per action (second)=0.022024.01 | 73.6 | — | 74.4 | 74 | — | — | |
| DfsrLite2025.10 | 73.6 | — | 74.4 | 74 | — | — | |
| DV2025.10 | 73.6 | — | 94 | 83.8 | — | — | |
| DiffuserLiteBackbone=R1, Runtime per action (second)=0.0142024.01 | 71.9 | — | 69.9 | 70.9 | — | — | |
| HD2025.10 | 71.7 | — | 73.3 | 72.5 | — | — | |
| HDMI2024.01 | 69.2 | — | — | — | — | — | |
| VIPRepresentation Strategy=VIP, MVL=false2026.02 | 69 | 25 | 63 | — | — | — | |
| BYOLRepresentation Strategy=BYOL, MVL=false2026.02 | 69 | 0 | 63 | — | — | — | |
| IDQL2025.10 | 66.5 | — | 66.7 | 66.6 | — | — | |
| IDQL*2025.10 | 66.5 | — | 66.7 | 66.6 | — | — | |
| DDRuntime per action (second)=2.4352024.01 | 65 | — | 57 | 61 | — | — | |
| DiffuserLiteBackbone=R2, Runtime per action (second)=0.0082024.01 | 64.8 | — | 71.4 | 68.1 | — | — | |
| DQL2025.10 | 62.6 | — | 60.5 | 61.6 | — | — | |
| MISA2022.10 | 56.6 | 70.2 | 45.7 | — | — | 172.5 | |
| SPQR-CQL-Min2024.01 | 55.6 | 47.9 | 54.2 | 52.6 | — | — | |
| DQL*2025.10 | 55.1 | — | 65.5 | 60.3 | — | — | |
| Diffuser2025.10 | 52.5 | — | 55.7 | 54.1 | — | — | |
| CQL2024.01 | 52.4 | — | 50.1 | 51.3 | — | — | |
| BC2022.10 | 51.5 | 65 | 38 | — | — | 154.5 | |
| BC2024.01 | 51.5 | — | 38 | 44.8 | — | — | |
| CQL-Min2024.01 | 51 | 43.8 | 49.8 | 48.2 | — | — | |
| CQL2022.10 | 51 | 43.8 | 49.8 | — | — | 144.6 | |
| IQL2022.10 | 51 | 62.5 | 46.3 | — | — | 159.8 | |
| IQL2024.01 | 51 | — | 46.3 | 48.7 | — | — | |
| 10%BC2022.10 | 50.9 | 7.2 | 66.8 | — | — | 124.9 | |
| DiffuserRuntime per action (second)=0.7492024.01 | 50 | — | 56.2 | 53.1 | — | — | |
| OneStep RL2022.10 | 47.6 | 57 | 53.1 | — | — | 157.7 | |
| BC2024.01 | 47.5 | 33.8 | 33.8 | 38.4 | — | — | |
| BC2025.10 | 47.5 | — | 33.8 | 40.7 | — | — | |
| BEAR2024.01 | 47.2 | 0 | 13.1 | 20.1 | — | — | |
| TD3+BC2022.10 | 46.2 | 16.8 | 22.4 | — | — | 85.4 | |
| SfBC2025.10 | 45.4 | — | 47.9 | 46.7 | — | — | |
| AWAC2022.10 | 22 | 39.3 | 36.6 | — | — | 97.9 | |
| SAC-Min2024.01 | 2.5 | 15 | 0 | 5.8 | — | — | |
| BC2024.01 | — | — | — | — | 33.8 | — | |
| BEAR2024.01 | — | — | — | — | 0 | — | |
| CQL-Min2024.01 | — | — | — | — | 43.8 | — | |
| IQL2024.01 | — | — | — | — | 62.5 | — | |
| SAC-Min2024.01 | — | — | — | — | 15 | — | |
| SPQRBase algorithm=CQL-Min2024.01 | — | — | — | — | 47.9 | — |