Open-loop Autonomous Driving on Navsim V1
100NCHuman
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| HumanBest-of-N sampling=false2026.03 | 100 | 100 | 87.5 | 100 | 99.9 | 94.8 | |
| Curious-VLABase Model=Qwen2.5-VL-3B, Sensors=1x C, Trajectory=Text Waypoint, Best-of-N sampling=true2026.03 | 99.5 | 99 | 91.8 | 99.3 | 98.4 | 94.8 | |
| DriveVLA-W0Base Model=Emu-3-8B, Sensors=1x C, Trajectory=Continuous, Best-of-N sampling=true2026.03 | 99.3 | 97.4 | 88.3 | 97 | 99.9 | 93 | |
| AutoVLABase Model=Qwen2.5-VL-3B, Sensors=3x C, Trajectory=Discrete Action, Best-of-N sampling=true2026.03 | 99.1 | 98.8 | 87.9 | 97.2 | 100 | 92.1 | |
| AdaThinkDriveBase Model=InternVL3-8B, Sensors=1x C, Trajectory=Text Waypoint, Best-of-N sampling=true2026.03 | 99.1 | 98.8 | 87.9 | 95.2 | 100 | 93 | |
| DriveVLA-W0Base Model=Emu-3-8B, Sensors=1x C, Trajectory=Continuous, Best-of-N sampling=false2026.03 | 98.7 | 99.1 | 93.3 | 95.3 | 99.3 | 90.2 | |
| ImagiDrive-SBase Model=InternVL2.5-4B, Sensors=1x C, Trajectory=Continuous, Best-of-N sampling=false2026.03 | 98.6 | 96.2 | 80.5 | 94.5 | 100 | 87.4 | |
| WoTESensors=3x C + L, Trajectory=Continuous, Best-of-N sampling=false2026.03 | 98.5 | 96.8 | 81.9 | 94.4 | 99.9 | 88.3 | |
| AutoVLABase Model=Qwen2.5-VL-3B, Sensors=3x C, Trajectory=Discrete Action, Best-of-N sampling=false2026.03 | 98.4 | 95.6 | 81.9 | 98 | 99.9 | 89.1 | |
| AdaThinkDriveBase Model=InternVL3-8B, Sensors=1x C, Trajectory=Text Waypoint, Best-of-N sampling=false2026.03 | 98.4 | 97.8 | 84.4 | 95.2 | 100 | 90.3 | |
| Curious-VLABase Model=Qwen2.5-VL-3B, Sensors=1x C, Trajectory=Text Waypoint, Best-of-N sampling=false2026.03 | 98.4 | 96.9 | 88.5 | 97.9 | 98.1 | 90.3 | |
| DiffusionDriveSensors=3x C + L, Trajectory=Continuous, Best-of-N sampling=false2026.03 | 98.2 | 96.2 | 82.2 | 94.7 | 100 | 88.1 | |
| ReCogDriveBase Model=InternVL2-8B, Sensors=3x C, Trajectory=Continuous, Best-of-N sampling=false2026.03 | 98.2 | 97.8 | 83.5 | 95.2 | 100 | 89.6 | |
| ImagiDrive-ABase Model=InternVL2.5-4B, Sensors=1x C, Trajectory=Continuous, Best-of-N sampling=false2026.03 | 98.1 | 96.2 | 80.1 | 94.4 | 100 | 86.9 | |
| Qwen2.5-VLBase Model=Qwen2.5-VL-7B, Sensors=1x C, Trajectory=Text Waypoint, Best-of-N sampling=false2026.03 | 97.8 | 92.1 | 78.3 | 92.8 | 100 | 83.3 | |
| UniADSensors=3x C + L, Trajectory=Continuous, Best-of-N sampling=false2026.03 | 97.7 | 92.8 | 79.2 | 92.8 | 100 | 84 | |
| InternVL2Base Model=InternVL2-8B, Sensors=1x C, Trajectory=Text Waypoint, Best-of-N sampling=false2026.03 | 97 | 92.4 | 78.9 | 91.8 | 100 | 83.3 | |
| Ego-MLPSensors=6x C, Trajectory=Continuous, Best-of-N sampling=false2026.03 | 93.1 | 78.3 | 63.2 | 84 | 100 | 66.4 |