Reinforcement Learning on DMC Quadruped
589.9Run ScoreFB-PbRL (Ours-FT)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| FB-PbRL (Ours-FT)fine-tuning=true2026.05 | 589.9 | 944.2 | 991.1 | 862.2 | 846.9 | |
| Ours-FT (FB-PbRL)Reward Specification=Preferences, Fine-tuning=true, Data Budget=2,000 preference pairs2026.05 | 575 | 889.4 | 990.1 | 850.9 | 826.3 | |
| FB-PbRL (Ours-T)test-time CPTS=true2026.05 | 503.4 | 606.1 | 886.1 | 658.1 | 663.4 | |
| OPRL2026.05 | 457.8 | 420.6 | 917.9 | 728.1 | 631.1 | |
| RLDPReward Specification=Ground-truth reward, Protocol=Zero-Shot RFRL, Data Budget=10,000 transitions2026.05 | 457.4 | 465.4 | 794.9 | 733.3 | 612.8 | |
| CLARIFY2026.05 | 444.7 | 350.6 | 901 | 715.4 | 602.9 | |
| PSMReward Specification=Ground-truth reward, Protocol=Zero-Shot RFRL, Data Budget=10,000 transitions2026.05 | 431.8 | 604 | 842.9 | 596.4 | 618.7 | |
| LaplaceReward Specification=Ground-truth reward, Protocol=Zero-Shot RFRL, Data Budget=10,000 transitions2026.05 | 413 | 494.6 | 854.5 | 642.8 | 601.2 | |
| FBReward Specification=Ground-truth reward, Protocol=Zero-Shot RFRL, Data Budget=10,000 transitions2026.05 | 397.4 | 529.8 | 720.2 | 599.1 | 561.7 | |
| OPPO2026.05 | 395.5 | 398.7 | 779.9 | 703 | 569.3 | |
| LIRE2026.05 | 246.8 | 284.3 | 523.7 | 459.9 | 378.7 | |
| HILPReward Specification=Ground-truth reward, Protocol=Zero-Shot RFRL, Data Budget=10,000 transitions2026.05 | 205.4 | 218.5 | 409.5 | 325.5 | 289.8 | |
| DPPO2026.05 | 204.5 | 194.4 | 458.2 | 379.1 | 309.1 |