Continuous Control on Hopper v5
3,732.5Average ReturnDBC
Evaluation results
| Method | Links | |
|---|---|---|
| DBCActor Backbone=SAC2026.02 | 3,732.5 | |
| TQCActor Backbone=SAC2026.02 | 3,704.6 | |
| SACActor Backbone=SAC2026.02 | 3,630.5 | |
| DSACActor Backbone=SAC2026.02 | 3,522.9 | |
| TRFP(ours)NFE=4 × 42026.04 | 3,507.5 | |
| TRFP(one-step)NFE=1 × 42026.04 | 3,440.6 | |
| VFActor Backbone=SAC2026.02 | 3,310.4 | |
| IQNActor Backbone=SAC2026.02 | 3,301.8 | |
| TD3NFE=12026.04 | 3,015.6 | |
| MaxEntDPNFE=20 × 102026.04 | 3,000.6 | |
| PDAEnvironment steps=1M, Seeds=10, Tests per epoch=102026.03 | 2,693.9 | |
| SDACNFE=20 × 322026.04 | 2,591.3 | |
| PPOEnvironment steps=1M, Seeds=10, Tests per epoch=102026.03 | 2,397.8 | |
| SACNFE=12026.04 | 2,220.6 | |
| VDActor Backbone=SAC2026.02 | 1,052.7 |