Autonomous Driving Planning on Navsim v1 (test)
100NCHuman
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| HumanVenue=-, Sensors=-2026.04 | 100 | 100 | 100 | 99.9 | 87.5 | 94.8 | |
| HumanCategory=Baselines2026.05 | 100 | 100 | 100 | 99.9 | 87.5 | 94.8 | |
| Human driver2026.06 | 100 | 100 | 100 | 99.9 | 87.5 | 94.8 | |
| DiffusionDrive w/ PaIR-DriveMethod type=Parallel IL+RL, best-of-N (N=6)=true2026.03 | 99.6 | 99.5 | 99.5 | 98.6 | 88.1 | 94 | |
| Transfuser w/ PaIR-DriveMethod type=Parallel IL+RL, best-of-N (N=6)=true2026.03 | 99.5 | 99.2 | 99.2 | 98.1 | 88 | 93.3 | |
| Centaur2026.06 | 99.5 | 98.9 | 98 | 100 | 85.9 | 92.6 | |
| DiffE2EInput=C&L2025.07 | 99.2 | 96.8 | 96.7 | 100 | 83.6 | 89.8 | |
| AutoVLA w/ GRPOMethod type=Sequential IL+RL, best-of-N (N=6)=true2026.03 | 99.1 | 97.1 | 97.1 | 100 | 87.6 | 92.1 | |
| Transfuser w/ PaIR-DriveMethod type=Parallel IL+RL, best-of-N (N=6)=false2026.03 | 99.1 | 96.1 | 98.2 | 93.1 | 88.1 | 89.7 | |
| DiffusionDrive w/ PaIR-DriveMethod type=Parallel IL+RL, best-of-N (N=6)=false2026.03 | 99.1 | 97.6 | 98.5 | 94.1 | 88.3 | 91.2 | |
| DriveWorld-VLAMethod Category=VLA-based Methods2026.05 | 99.1 | 98.2 | 96.1 | 100 | 85.9 | 91.3 | |
| RAP-DINO2026.06 | 99.1 | 98.9 | 96.7 | 100 | 90.3 | 93.8 | |
| DrivoR2026.06 | 99.1 | 99.2 | 96.9 | 100 | 91.6 | 94.6 | |
| DriveLaWMethod Category=World-Model-based Methods2026.05 | 99 | 97.1 | 96.7 | 100 | 81.3 | 89.1 | |
| DriveLaWVenue=CVPR’26, Input=C, WM=✓2026.05 | 99 | 97.1 | 96.7 | 100 | 81.3 | 89.1 | |
| CoPhyVenue=–, Input=C+L, WM=✓2026.05 | 99 | 98.2 | 96.8 | 100 | 85.3 | 91.4 | |
| GTRS (V2-99)Model Version=V2-992026.06 | 99 | 98.9 | 95.9 | 100 | 82.8 | 90.4 | |
| iPad + TOADHeuristic=TOAD2026.06 | 99 | 99.3 | 96.4 | 100 | 89.1 | 93.4 | |
| RAP-DINO + TOADHeuristic=TOAD2026.06 | 99 | 99.2 | 96.6 | 100 | 90.1 | 93.9 | |
| DrivoR + TOADHeuristic=TOAD2026.06 | 99 | 99.3 | 96.8 | 100 | 91.8 | 94.7 | |
| DrivingGPTVenue=ICCV’25, Sensors=1×C2026.04 | 98.9 | 90.7 | 94.9 | 95.6 | 79.7 | 82.4 | |
| DrivingGPT2026.06 | 98.9 | 90.7 | 94.9 | 95.6 | 79.7 | 82.4 | |
| GTRS (V2-99) + TOADModel Version=V2-99, Heuristic=TOAD2026.06 | 98.9 | 98.1 | 96.3 | 100 | 84.8 | 90.9 | |
| DriveFutureMethod Category=World-Model-based Methods2026.05 | 98.8 | 99.1 | 95.4 | 100 | 84.2 | 90.7 | |
| ZTRS (V2-99)Model Version=V2-992026.06 | 98.8 | 99.5 | 95.1 | 100 | 74.7 | 86.9 | |
| DriveVLA-W0Venue=ICLR’26, Sensors=1×C2026.04 | 98.7 | 96.2 | 95.5 | 100 | 82.2 | 88.4 | |
| DriveVLA-W0Method Category=VLA-based Methods2026.05 | 98.7 | 99.1 | 95.3 | 99.3 | 83.3 | 90.2 | |
| DriveVLA-W02026.06 | 98.7 | 99.1 | 95.3 | 99.3 | 83.3 | 90.2 | |
| Hydra-MDP + TOADHeuristic=TOAD2026.06 | 98.7 | 98 | 95.4 | 100 | 87.2 | 91.4 | |
| ipadImg. Backbone=ResNet-342026.03 | 98.6 | 98.3 | 94.9 | 100 | 88 | 91.7 | |
| PWMVenue=NeurIPS’25, Sensors=1×C2026.04 | 98.6 | 95.9 | 95.4 | 100 | 81.8 | 88.1 | |
| Hydra-MDP++2026.06 | 98.6 | 98.6 | 95.1 | 100 | 85.7 | 91 | |
| DriveSuprim2026.06 | 98.6 | 98.6 | 95.5 | 100 | 91.3 | 93.5 | |
| iPad2026.06 | 98.6 | 98.3 | 94.9 | 100 | 88 | 91.7 | |
| WoTEMethod type=IL, best-of-N (N=6)=false2026.03 | 98.5 | 96.8 | 94.9 | 99.9 | 81.9 | 88.3 | |
| DriveDPO w/ DPOMethod type=Sequential IL+RL, best-of-N (N=6)=false2026.03 | 98.5 | 98.1 | 94.8 | 99.9 | 84.3 | 90 | |
| WoTEImg. Backbone=ResNet-342026.03 | 98.5 | 96.8 | 94.9 | 99.9 | 81.9 | 88.3 | |
| DIVERImg. Backbone=ResNet-342026.03 | 98.5 | 96.5 | 94.9 | 100 | 82.6 | 88.3 | |
| SparseDriveV2Img. Backbone=ResNet-342026.03 | 98.5 | 98.4 | 95 | 99.9 | 88.6 | 92 | |
| WoTEVenue=ICCV’25, Sensors=3×C + L2026.04 | 98.5 | 96.8 | 94.4 | 99.9 | 81.9 | 88.3 | |
| DIVERMethod Category=E2E-based Methods2026.05 | 98.5 | 96.5 | 94.9 | 100 | 82.6 | 88.3 | |
| WoTEMethod Category=World-Model-based Methods2026.05 | 98.5 | 96.8 | 94.9 | 99.9 | 81.9 | 88.3 | |
| DIVERInput=C&L2025.07 | 98.5 | 96.5 | 94.9 | 100 | 82.6 | 88.3 | |
| WoTEVenue=ICCV’25, Input=C+L, WM=✓2026.05 | 98.5 | 96.8 | 94.9 | 99.9 | 81.9 | 88.3 | |
| DriveDPOVenue=NeurIPS’25, Input=C+L, WM=✗2026.05 | 98.5 | 98.1 | 94.8 | 99.9 | 84.3 | 90 | |
| DIVER2026.06 | 98.5 | 96.5 | 94.9 | 100 | 82.6 | 88.3 | |
| Hydra-MDP2026.06 | 98.5 | 98.7 | 94.9 | 100 | 85.5 | 90.9 | |
| Hydra-MDPImg. Backbone=V2-992026.03 | 98.4 | 97.8 | 93.9 | 100 | 86.5 | 90.3 | |
| GoalFlowImg. Backbone=V2-992026.03 | 98.4 | 98.3 | 94.6 | 100 | 85 | 90.3 | |
| AutoVLAVenue=NeurIPS’25, Sensors=3×C2026.04 | 98.4 | 95.6 | 98 | 99.9 | 81.9 | 89.1 | |
| DriveDreamer-PolicyVenue=-, Sensors=3×C2026.04 | 98.4 | 97.1 | 95.1 | 100 | 83.5 | 89.2 | |
| GoalFlowMethod Category=E2E-based Methods, Row=22026.05 | 98.4 | 98.3 | 94.6 | 100 | 85 | 90.3 | |
| AutoVLAMethod Category=VLA-based Methods2026.05 | 98.4 | 95.6 | 98 | 99.9 | 81.9 | 89.1 | |
| DiffRefinerInput=C2025.07 | 98.4 | 97.4 | 95.3 | 100 | 83.4 | 89.4 | |
| GoalFlowInput=C&L2025.07 | 98.4 | 98.3 | 94.6 | 100 | 85 | 90.3 | |
| AutoVLA2026.06 | 98.4 | 95.6 | 98 | 99.9 | 81.9 | 89.1 | |
| ZTRS (V2-99) + TOADModel Version=V2-99, Heuristic=TOAD2026.06 | 98.4 | 98.6 | 94.3 | 100 | 82.1 | 89 | |
| ReCogDrive w/o RLMethod type=IL, best-of-N (N=6)=false2026.03 | 98.3 | 95.1 | 94.3 | 100 | 81.1 | 86.8 | |
| ARTEMISMethod type=IL, best-of-N (N=6)=false2026.03 | 98.3 | 95.1 | 94.3 | 100 | 81.4 | 87 | |
| GoalFlowImg. Backbone=ResNet-342026.03 | 98.3 | 93.8 | 94.3 | 100 | 79.8 | 85.7 | |
| Hydra-MDPImg. Backbone=ResNet-342026.03 | 98.3 | 96 | 94.6 | 100 | 78.7 | 86.5 | |
| ARTEMISImg. Backbone=ResNet-342026.03 | 98.3 | 95.1 | 94.3 | 100 | 81.4 | 87 | |
| DiffusionDriveV2Img. Backbone=ResNet-342026.03 | 98.3 | 97.9 | 94.8 | 99.9 | 87.5 | 91.2 | |
| GoalFlowMethod Category=E2E-based Methods, Row=12026.05 | 98.3 | 93.8 | 94.3 | 100 | 79.8 | 85.7 | |
| Hydra-MDPMethod Category=E2E-based Methods2026.05 | 98.3 | 96 | 94.6 | 100 | 78.7 | 86.5 | |
| ARTEMISMethod Category=E2E-based Methods2026.05 | 98.3 | 95.1 | 94.3 | 100 | 81.4 | 87 | |
| Hydra-MDPInput=C&L2025.07 | 98.3 | 96 | 94.6 | 100 | 78.7 | 86.5 | |
| ARTEMISVenue=RAL’25, Input=C+L, WM=✗2026.05 | 98.3 | 95.1 | 94.3 | 100 | 81.4 | 87 | |
| GoalFlowVenue=CVPR’25, Input=C+L, WM=✗2026.05 | 98.3 | 93.8 | 94.3 | 100 | 79.8 | 85.7 | |
| VADv2Venue=ICLR’26, Input=C, WM=✗2026.05 | 98.3 | 97.4 | 95.7 | 100 | 82.3 | 89.3 | |
| GoalFlowBackbone=ResNet, Feature encoding=256-dimensional2025.12 | 98.2 | 96.4 | 93.8 | 99.4 | 82.6 | 87.9 | |
| Diff.2025.12 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | |
| Mimir2025.12 | 98.2 | 97.5 | 94.6 | 100 | 83.6 | 89.3 | |
| DiffusionDriveMethod type=IL, best-of-N (N=6)=false2026.03 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | |
| ReCogDrive w/ GRPOMethod type=Sequential IL+RL, best-of-N (N=6)=false2026.03 | 98.2 | 97.8 | 95.2 | 99.8 | 83.5 | 89.6 | |
| DiffusionDriveImg. Backbone=ResNet-342026.03 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | |
| DiffusionDriveVenue=CVPR’25, Sensors=3×C + L2026.04 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | |
| FSDriveVenue=NeurIPS’25, Sensors=3×C2026.04 | 98.2 | 93.8 | 93.3 | 99.9 | 80.1 | 85.1 | |
| DiffusionDriveMethod Category=E2E-based Methods2026.05 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | |
| RecogdriveMethod Category=VLA-based Methods2026.05 | 98.2 | 97.8 | 95.2 | 99.8 | 83.5 | 89.6 | |
| DiffusionDriveInput=C&L2025.07 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | |
| BridgeDriveInput=C&L2025.07 | 98.2 | 96.1 | 94.5 | 100 | 82.3 | 88 | |
| DiffusionDriveVenue=CVPR’25, Input=C+L, WM=✗2026.05 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | |
| DiffusionDrive2026.06 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | |
| FUMPImg. Backbone=ResNet-342026.03 | 98.1 | 96.2 | 94.2 | 100 | 82 | 87.8 | |
| Recogdrive*Venue=ICLR’26, Sensors=3×C2026.04 | 98.1 | 94.7 | 94.2 | 100 | 80.9 | 86.5 | |
| VAD-v22026.06 | 98.1 | 94.8 | 94.3 | 100 | 80.6 | 86.2 | |
| PRIX2026.06 | 98.1 | 96.3 | 94.1 | 100 | 82.3 | 87.8 | |
| Transfuser w/ GRPOMethod type=Sequential IL+RL, best-of-N (N=6)=false2026.03 | 98 | 94.7 | 96.6 | 100 | 88.5 | 87.9 | |
| DRAMAImg. Backbone=ResNet-342026.03 | 98 | 93.1 | 94.8 | 100 | 80.1 | 85.5 | |
| DRAMAMethod Category=E2E-based Methods2026.05 | 98 | 93.1 | 94.8 | 100 | 80.1 | 85.5 | |
| DRAMAVenue=arXiv’24, Input=C+L, WM=✗2026.05 | 98 | 93.1 | 94.8 | 100 | 80.1 | 85.5 | |
| ResADVenue=CVPR’26, Input=C+L, WM=✗2026.05 | 98 | 97.5 | 94.1 | 100 | 83.3 | 88.8 | |
| DRAMA2026.06 | 98 | 93.1 | 94.8 | 100 | 80.1 | 85.5 | |
| PARA-Drive2025.12 | 97.9 | 92.4 | 93 | 99.8 | 79.3 | 84 | |
| DriveDPO w/o RLMethod type=IL, best-of-N (N=6)=false2026.03 | 97.9 | 97.3 | 93.6 | 100 | 84 | 88.8 | |
| PARA-DriveImg. Backbone=ResNet-342026.03 | 97.9 | 92.4 | 93 | 99.8 | 79.3 | 84 | |
| PARA-DriveVenue=CVPR’24, Sensors=6×C2026.04 | 97.9 | 92.4 | 93 | 99.8 | 79.3 | 84 | |
| EponaVenue=ICCV’25, Sensors=3×C2026.04 | 97.9 | 95.1 | 93.8 | 99.9 | 80.4 | 86.2 | |
| PARA-DriveMethod Category=E2E-based Methods2026.05 | 97.9 | 92.4 | 93 | 99.8 | 79.3 | 84 |