Autonomous Driving Planning on NAVSIM v1
100NCHuman
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| HumanSensors=N/A2026.02 | 100 | 100 | 100 | 99.9 | 87.5 | 94.8 | — | |
| Human2026.03 | 100 | 100 | 100 | 99.9 | 87.5 | 94.8 | — | |
| HumanInput=—2026.04 | 100 | 100 | 100 | 99.9 | 87.5 | 94.8 | — | |
| Human driver2026.05 | 100 | 100 | 100 | 99.9 | 87.5 | 94.8 | — | |
| HumanSize=-, Sensors=–2026.05 | 100 | 100 | 100 | 99.9 | 87.5 | 94.8 | — | |
| DriveVLA-W0Input=SC, Best-of-N strategy=true2026.04 | 99.9 | 97.4 | 97 | 99.9 | 88.3 | 93 | — | |
| Centaur2026.05 | 99.5 | 98.9 | 98 | 100 | 85.9 | 92.6 | — | |
| ExploreVLAInput=SC, Best-of-N strategy=true2026.04 | 99.4 | 98.9 | 98.3 | 99.7 | 88.3 | 93.7 | — | |
| DriveVLA-W0Sensors=1xC, best-of-6 sampling=true2026.02 | 99.3 | 97.4 | 97 | 99.9 | 88.3 | 93 | — | |
| DriveFineSensors=1xC, best-of-6 sampling=true2026.02 | 99.3 | 99.2 | 97.9 | 100 | 89.1 | 94.2 | — | |
| DriveVLA-W0Method category=VLA-based Methods, Sensors=1x Cam, Best-of-N strategy=true2026.03 | 99.3 | 97.4 | 97 | 99.9 | 88.3 | 93 | — | |
| VegaSensors=1x Cam, Best-of-N strategy=true2026.03 | 99.2 | 96.6 | 96.9 | 100 | 83.4 | 89.8 | — | |
| CentaurInput=MC+L, Best-of-N strategy=false2026.04 | 99.2 | 98.7 | 97.2 | 99.9 | 86 | 92.1 | — | |
| AutoVLASensors=3xC, best-of-6 sampling=true2026.02 | 99.1 | 97.1 | 97.1 | 100 | 87.6 | 92.1 | — | |
| AdaThinkDriveSensors=1xC, best-of-6 sampling=true2026.02 | 99.1 | 98.8 | 97.2 | 100 | 87.9 | 93 | — | |
| AutoVLAMethod category=VLA-based Methods, Sensors=3x Cam, Best-of-N strategy=true2026.03 | 99.1 | 97.1 | 97.1 | 100 | 87.6 | 92.1 | — | |
| Hydra-MDPInput=MC+L, Best-of-N strategy=false2026.04 | 99.1 | 98.3 | 96.6 | 100 | 85.2 | 91.3 | — | |
| AutoVLAInput=MC, Best-of-N strategy=true2026.04 | 99.1 | 97.1 | 97.1 | 99.9 | 87.6 | 92.1 | — | |
| Clover2026.05 | 99.1 | 99 | 96.9 | 100 | 91.7 | 94.5 | — | |
| Ours (SFT)Size=0.5B, Sensors=C, Training=SFT2026.05 | 99.1 | 96.2 | 99 | 100 | 83.4 | 90.3 | — | |
| Ours†Size=0.5B, Sensors=C, Training=SFT→GRPO2026.05 | 99.1 | 97.8 | 99.1 | 99.9 | 85.4 | 92.2 | — | |
| DrivoR2026.05 | 99 | 98.9 | 96.7 | 100 | 90 | 93.7 | — | |
| ELF-VLA-8BImage=true, Lidar=false2026.03 | 98.9 | 98.1 | 96 | — | 85.3 | 91 | 100 | |
| VegaSensors=1x Cam, Best-of-N strategy=false2026.03 | 98.9 | 95.3 | 96.1 | 100 | 81.6 | 87.9 | — | |
| DrivingGPTInput=SC, Best-of-N strategy=false2026.04 | 98.9 | 90.7 | 94.9 | 95.6 | 79.9 | 82.4 | — | |
| PWMInput=SC, Best-of-N strategy=false2026.04 | 98.9 | 95.8 | 95.9 | 100 | 81.5 | 88.1 | — | |
| DrivingGPT2026.05 | 98.9 | 90.7 | 94.9 | 95.6 | 79.7 | 82.4 | — | |
| DriveFineSensors=1xC, score-based reinforcement training=true2026.02 | 98.8 | 98.6 | 96.2 | 100 | 86.9 | 91.8 | — | |
| ExploreVLAInput=SC, Best-of-N strategy=false2026.04 | 98.8 | 98.4 | 96.5 | 99.9 | 83.5 | 90.4 | — | |
| DriveFineSize=8B, Sensors=C2026.05 | 98.8 | 98.6 | 96.2 | 100 | 86.9 | 91.8 | — | |
| DriveVLA-W0Sensors=1xC2026.02 | 98.7 | 99.1 | 95.3 | 99.3 | 83.3 | 90.2 | — | |
| DriveVLA-W0-3BImage=true, Lidar=false2026.03 | 98.7 | 99.1 | 95.3 | — | 83.3 | 90.3 | 99.3 | |
| DriveVLA-W0-7BImage=true, Lidar=false2026.03 | 98.7 | 99.1 | 94.9 | — | 83.3 | 90.2 | 99.3 | |
| LaST-VLA-8B (SFT)Image=true, Lidar=false, Training Protocol=SFT2026.03 | 98.7 | 95.4 | 95.6 | — | 80.5 | 87.3 | 100 | |
| LaST-VLA-8B (RL)Image=true, Lidar=false, Training Protocol=RL2026.03 | 98.7 | 97.9 | 95.6 | — | 86.8 | 91.3 | 100 | |
| DriveVLA-W0Input=SC, Best-of-N strategy=false2026.04 | 98.7 | 99.1 | 97.1 | 100 | 87.6 | 90.2 | — | |
| DriveVLA-W02026.05 | 98.7 | 99.1 | 95.3 | 99.3 | 83.3 | 90.2 | — | |
| DriveVLA-W0Size=7B, Sensors=C2026.05 | 98.7 | 99.1 | 95.3 | 99.3 | 83.3 | 90.2 | — | |
| DriveFineSensors=1xC2026.02 | 98.6 | 97.9 | 95.2 | 99.9 | 85.5 | 90.7 | — | |
| LaST-VLA-2B (RL)Image=true, Lidar=false, Training Protocol=RL2026.03 | 98.6 | 97.7 | 95.6 | — | 86.5 | 91.1 | 100 | |
| DriveSuprimInput=MC+L, Best-of-N strategy=false2026.04 | 98.6 | 98.6 | 95.5 | 100 | 91.3 | 93.5 | — | |
| Hydra-MDP++2026.05 | 98.6 | 98.6 | 95.1 | 100 | 85.7 | 91 | — | |
| iPad2026.05 | 98.6 | 98.3 | 94.9 | 100 | 88 | 91.7 | — | |
| DriveSuprim2026.05 | 98.6 | 98.6 | 95.5 | 100 | 91.3 | 93.5 | — | |
| WoTESensors=3xC+L2026.02 | 98.5 | 96.8 | 94.4 | 99.9 | 81.9 | 88.3 | — | |
| WoTEImage=true, Lidar=true2026.03 | 98.5 | 96.8 | 94.9 | — | 81.9 | 88.3 | 99.9 | |
| InternVL3-8B-SFTImage=true, Lidar=false2026.03 | 98.5 | 95.5 | 95.3 | — | 81.2 | 87.4 | 100 | |
| InternVL3-8B-RLImage=true, Lidar=false2026.03 | 98.5 | 96.7 | 95.4 | — | 83.2 | 89 | 100 | |
| WoTEImage=true, Lidar=true2026.03 | 98.5 | 96.8 | 94.9 | — | 81.9 | 88.3 | 99.9 | |
| LaST-VLA-2B (SFT)Image=true, Lidar=false, Training Protocol=SFT2026.03 | 98.5 | 95.2 | 95.6 | — | 80.4 | 87 | 100 | |
| WoTEMethod category=BEV-based Methods, Sensors=3x Cam + L2026.03 | 98.5 | 96.8 | 94.4 | 99.9 | 81.9 | 88.3 | — | |
| WoTEInput=C+L2026.04 | 98.5 | 96.8 | 94.9 | 99.9 | 81.9 | 88.3 | — | |
| DIVER2026.05 | 98.5 | 96.5 | 94.9 | 100 | 82.6 | 88.3 | — | |
| WoTESize=0.05B, Sensors=C+L2026.05 | 98.5 | 96.8 | 94.4 | 99.9 | 81.9 | 88.3 | — | |
| AutoVLASensors=3xC2026.02 | 98.4 | 95.6 | 98 | 99.9 | 81.9 | 89.1 | — | |
| AdaThinkDriveSensors=1xC2026.02 | 98.4 | 97.8 | 95.2 | 100 | 84.4 | 90.3 | — | |
| AutoVLA-3BImage=true, Lidar=false2026.03 | 98.4 | 95.6 | 98 | — | 81.9 | 89.1 | 100 | |
| GoalFlowImage=true, Lidar=true2026.03 | 98.4 | 98.3 | 94.6 | — | 85 | 90.3 | 100 | |
| AdaThinkDrive-8BImage=true, Lidar=false, Textual CoT=true2026.03 | 98.4 | 97.8 | 95.2 | — | 84.4 | 90.3 | 100 | |
| AutoVLAMethod category=VLA-based Methods, Sensors=3x Cam, Best-of-N strategy=false2026.03 | 98.4 | 95.6 | 98 | 99.9 | 81.9 | 89.1 | — | |
| AutoVLAInput=MC, Best-of-N strategy=false2026.04 | 98.4 | 95.6 | 98 | 99.9 | 81.9 | 89.1 | — | |
| AutoVLA2026.05 | 98.4 | 95.6 | 98 | 99.9 | 81.9 | 89.1 | — | |
| AutoVLASize=3B, Sensors=C2026.05 | 98.4 | 95.6 | 98 | 99.9 | 81.9 | 89.1 | — | |
| AdaThinkDriveSize=2B, Sensors=C2026.05 | 98.4 | 97.8 | 95.2 | 100 | 84.4 | 90.3 | — | |
| Hydra-MDPSensors=3xC+L2026.02 | 98.3 | 96 | 94.6 | 100 | 78.7 | 86.5 | — | |
| InternVL3-8B (SFT)Image=true, Lidar=false, Textual CoT=true, Training Protocol=SFT2026.03 | 98.3 | 92.3 | 94.7 | — | 77.2 | 83.9 | 100 | |
| InternVL3-8B (RL)Image=true, Lidar=false, Textual CoT=true, Training Protocol=RL2026.03 | 98.3 | 94 | 94.7 | — | 83 | 87.2 | 100 | |
| Hydra-MDPMethod category=BEV-based Methods, Sensors=3x Cam + L2026.03 | 98.3 | 96 | 94.6 | 100 | 78.7 | 86.5 | — | |
| GoalFlowInput=C+L2026.04 | 98.3 | 93.8 | 94.3 | 100 | 79.8 | 85.7 | — | |
| Hydra-MDPInput=C+L2026.04 | 98.3 | 96 | 94.6 | 100 | 78.7 | 86.5 | — | |
| Hydra-MDPSize=0.05B, Sensors=C+L2026.05 | 98.3 | 96 | 94.6 | 100 | 78.7 | 86.5 | — | |
| DiffusionDriveSensors=3xC+L2026.02 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | — | |
| ReCogDriveSensors=1xC2026.02 | 98.2 | 97.8 | 95.2 | 99.8 | 83.5 | 89.6 | — | |
| DiffusionDriveImage=true, Lidar=true2026.03 | 98.2 | 96.2 | 94.7 | — | 82.2 | 88.1 | 100 | |
| DiffusionDriveImage=true, Lidar=true2026.03 | 98.2 | 96.2 | 94.7 | — | 82.2 | 88.1 | 100 | |
| DiffusionDriveMethod category=BEV-based Methods, Sensors=3x Cam + L2026.03 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | — | |
| ReCogDriveMethod category=VLA-based Methods, Sensors=3x Cam2026.03 | 98.2 | 97.8 | 95.2 | 99.8 | 83.5 | 89.6 | — | |
| DRAMAInput=MC+L, Best-of-N strategy=false2026.04 | 98.2 | 95.2 | 94.2 | 100 | 81.3 | 86.9 | — | |
| FSDriveInput=SC, Best-of-N strategy=false2026.04 | 98.2 | 93.8 | 93.3 | 99.9 | 80.1 | 85.1 | — | |
| DiffusionDriveInput=C+L2026.04 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | — | |
| HADInput=C+L2026.04 | 98.2 | 97.3 | 97.5 | 100 | 87.4 | 90.2 | — | |
| DiffusionDrive2026.05 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | — | |
| DiffusionDriveSize=0.05B, Sensors=C+L2026.05 | 98.2 | 96.2 | 94.7 | 100 | 82.2 | 88.1 | — | |
| ReCogDriveSize=2B, Sensors=C2026.05 | 98.2 | 97.8 | 95.2 | 99.8 | 83.5 | 89.6 | — | |
| Hydra-NeXtImage=true, Lidar=false2026.03 | 98.1 | 97.7 | 94.6 | — | 81.8 | 88.6 | 100 | |
| HAD-LInput=C2026.04 | 98.1 | 97.2 | 97.3 | 100 | 87.2 | 89.9 | — | |
| VAD-v22026.05 | 98.1 | 94.8 | 94.3 | 100 | 80.6 | 86.2 | — | |
| PRIX2026.05 | 98.1 | 96.3 | 94.1 | 100 | 82.3 | 87.8 | — | |
| SUPER-ADSize=0.05B, Sensors=C2026.05 | 98.1 | 97 | 93.9 | 99.8 | 81.5 | 87.7 | — | |
| DRAMAInput=C+L2026.04 | 98 | 93.1 | 94.8 | 100 | 80.1 | 85.5 | — | |
| DRAMA2026.05 | 98 | 93.1 | 94.8 | 100 | 80.1 | 85.5 | — | |
| Recogdrive-2BImage=true, Lidar=false2026.03 | 97.9 | 97.3 | 94.9 | — | 87.3 | 90.8 | 100 | |
| PARA-DriveMethod category=BEV-based Methods, Sensors=6x Cam2026.03 | 97.9 | 92.4 | 93 | 99.8 | 79.3 | 84 | — | |
| VADv2Input=C2026.04 | 97.9 | 91.7 | 92.9 | 100 | 77.6 | 83 | — | |
| PARA-Drive2026.05 | 97.9 | 92.4 | 93 | 99.8 | 79.3 | 84 | — | |
| ReCogDrive2026.05 | 97.9 | 97.3 | 94.9 | 100 | 87.3 | 90.8 | — | |
| UniADSensors=6xC2026.02 | 97.8 | 91.9 | 92.9 | 100 | 78.8 | 83.4 | — | |
| UniADImage=true, Lidar=false2026.03 | 97.8 | 91.9 | 92.9 | — | 78.8 | 83.4 | 100 | |
| UniADImage=true, Lidar=false2026.03 | 97.8 | 91.9 | 92.9 | — | 78.8 | 83.4 | 100 | |
| UniADMethod category=BEV-based Methods, Sensors=6x Cam2026.03 | 97.8 | 91.9 | 92.9 | 100 | 78.8 | 83.4 | — |