Closed-loop Planning on Bench2Drive (test)
86.77Driving ScoreHiP-AD
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| HiP-AD2025.03 | 86.77 | 69.09 | 203.12 | 19.36 | |
| DiffADInput=Ego State + 6 Cameras, Expert Feature Distillation=true2025.04 | 67.92 | 38.64 | — | — | |
| DriveAdapterInput=Ego State + 6 Cameras, Expert Feature Distillation=true2025.04 | 64.22 | 33.08 | — | — | |
| DriveAdapterexpert feature distillation=true2025.03 | 64.22 | 33.08 | 70.22 | 16.01 | |
| Drive Transformer-Large2025.03 | 63.46 | 35.01 | 100.64 | 20.78 | |
| Think2TwiceInput=Ego State + 6 Cameras, Expert Feature Distillation=true2025.04 | 62.44 | 31.23 | — | — | |
| ThinkTwiceexpert feature distillation=true2025.03 | 62.44 | 31.23 | 69.33 | 16.22 | |
| WOTEInput=Ego State + 6 Cameras, Expert Feature Distillation=true2025.04 | 61.71 | 31.36 | — | — | |
| TCP-trajexpert feature distillation=true2025.03 | 59.9 | 30 | 76.54 | 18.08 | |
| DiFSD2025.03 | 52.02 | 21 | 178.3 | — | |
| UniAD-BaseInput=Ego State + 6 Cameras, Expert Feature Distillation=false2025.04 | 45.81 | 16.36 | — | — | |
| UniAD-Base2025.03 | 45.81 | 16.36 | 129.21 | 43.58 | |
| FUMPInput=Ego State + 6 Cameras, Expert Feature Distillation=false2025.04 | 45.67 | 16.36 | — | — | |
| GenAD2025.03 | 44.81 | 15.9 | — | — | |
| SparseDrive2025.03 | 44.54 | 16.71 | 170.21 | 48.63 | |
| VADInput=Ego State + 6 Cameras, Expert Feature Distillation=false2025.04 | 42.35 | 15 | — | — | |
| VAD2025.03 | 42.35 | 15 | 157.94 | 46.01 | |
| UniAD-TinyInput=Ego State + 6 Cameras, Expert Feature Distillation=false2025.04 | 40.73 | 13.18 | — | — | |
| UniAD-Tiny2025.03 | 40.73 | 13.18 | 123.92 | 47.04 | |
| AD-MLPInput=Ego State, Expert Feature Distillation=false2025.04 | 18.05 | 0 | — | — | |
| AD-MLP2025.03 | 18.05 | 0 | 48.45 | 22.63 |