Locomotion on D4RL HalfCheetah
88.85ReturnOurs
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| OursProtocol=one-step distillation, IQL-based candidate scoring2026.05 | 88.85 | 0.82 | 1.34 | 84.88 | |
| Ours (classifier)Protocol=one-step distillation, teacher-side classifier2026.05 | 85.99 | 1.83 | 0.81 | 84.17 | |
| TeacherProtocol=20-step reverse diffusion with classifier-based re-ranking2026.05 | 85.18 | 116.51 | 1.14 | 82.29 | |
| KDMProtocol=one-step comparison, IQL-based candidate scoring2026.05 | 66.18 | 0.34 | 3.49 | 65.71 | |
| KDM-FProtocol=one-step comparison, IQL-based candidate scoring2026.05 | 48.52 | 0.86 | 2.16 | 46.59 | |
| CTProtocol=one-step comparison, IQL-based candidate scoring2026.05 | 40.89 | 1.08 | 1.98 | 36.63 | |
| CDProtocol=one-step comparison, IQL-based candidate scoring2026.05 | 40.27 | 1.08 | 2.77 | 32.17 |