Trajectory Planning on nuScenes (val)
0.07Collision Rate (Avg)GuideFlow
Evaluation results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GuideFlowInput=Camera, Backbone=ResNet50, FPS=3.6 (RTX4090)2025.11 | 0.07 | — | — | — | — | 0 | 0.02 | 0.18 | — | — | — | — | — | — | |
| SparseDrive*Input=Camera, Backbone=ResNet50, FPS=9.0 (RTX4090)2025.11 | 0.08 | — | — | — | — | 0.01 | 0.05 | 0.18 | — | — | — | — | — | — | |
| MomADInput=Camera, Backbone=ResNet50, FPS=7.8 (RTX4090)2025.11 | 0.09 | — | — | — | — | 0.01 | 0.05 | 0.22 | — | — | — | — | — | — | |
| VADInput=Camera, Backbone=ResNet502025.11 | 0.21 | — | — | — | — | 0.03 | 0.19 | 0.43 | — | — | — | — | — | — | |
| UniADInput=Camera, Backbone=ResNet101, FPS=1.8 (A100)2025.11 | 0.61 | — | — | — | — | 0.62 | 0.58 | 0.63 | — | — | — | — | — | — | |
| WorldRFT (with RFT)Training Approach=SS-L & RL, Ego status=true2025.12 | 4 | 0.15 | 0.3 | 0.56 | 0.33 | 0 | 0 | 12 | — | — | — | — | — | — | |
| FlowAD_DiffusionDriveBackbone=ResNet502026.03 | 4 | — | — | — | 0.54 | — | — | — | — | — | — | — | — | 1.05 | |
| FlowAD_SparseDriveBackbone=ResNet50, ego status=true2026.03 | 4 | — | — | — | 0.25 | — | — | — | — | — | — | — | — | 0.75 | |
| DiffusionDrive-HATStage=2nd2025.12 | 4.2 | — | — | — | — | — | — | — | 0.58 | — | — | — | — | — | |
| WorldRFT (with RFT)Training Approach=SS-L & RL, Ego status=false2025.12 | 5 | 0.22 | 0.44 | 0.77 | 0.48 | 0 | 0 | 16 | — | — | — | — | — | — | |
| FlowAD_SparseDriveBackbone=ResNet1012026.03 | 5 | — | — | — | 0.52 | — | — | — | — | — | — | — | — | 0.91 | |
| SparseDrive-B2024.05 | 6 | 0.29 | 0.55 | 0.91 | 0.58 | 1 | 2 | 13 | — | — | — | — | — | — | |
| DIMA (UniAD)Using Ego status=true2025.01 | 6 | 0.19 | 0.5 | 1.08 | 0.59 | — | — | — | — | — | — | — | — | — | |
| DIMA (UniAD)Using Ego status=true2025.01 | 6 | 0.3 | 0.82 | 1.63 | 0.92 | — | — | — | — | — | — | — | — | — | |
| SSRStage=-2025.12 | 6 | — | — | — | — | — | — | — | 0.39 | — | — | — | — | — | |
| FlowAD_SparseDriveBackbone=ResNet502026.03 | 6 | — | — | — | 0.56 | — | — | — | — | — | — | — | — | 1.03 | |
| SparseDriveBackbone=ResNet1012026.03 | 6 | — | — | — | 0.58 | — | — | — | — | — | — | — | — | 2.3 | |
| SSRBackbone=ResNet50, ego status=true2026.03 | 6 | — | — | — | 0.39 | — | — | — | — | — | — | — | — | 1.91 | |
| UncADBackbone=ResNet502026.03 | 7 | — | — | — | 0.6 | — | — | — | — | — | — | — | — | — | |
| SparseDrive-S2024.05 | 8 | 0.29 | 0.58 | 0.96 | 0.61 | 1 | 5 | 18 | — | — | — | — | — | — | |
| DiffusionDriveStage=2nd2025.12 | 8 | — | — | — | — | — | — | — | 0.57 | — | — | — | — | — | |
| DiffusionDriveTraining Approach=P-IL, Ego status=false2025.12 | 8 | 0.27 | 0.54 | 0.9 | 0.57 | 3 | 5 | 16 | — | — | — | — | — | — | |
| SparseDriveInput=Camera, Backbone=ResNet50, Reproduction Status=Official Checkpoint, FPS=9.0, Hardware=RTX40902026.03 | 8 | 0.29 | 0.58 | 0.96 | 0.61 | 1 | 5 | 18 | — | 0.3 | 0.57 | 0.85 | 0.57 | — | |
| Ours (Neuro-Symbolic)Input=Camera, Backbone=ResNet50, FPS=10.0, Hardware=H2002026.03 | 8 | 0.27 | 0.54 | 0.9 | 0.57 | 1 | 5 | 17 | — | 0.26 | 0.48 | 0.66 | 0.47 | — | |
| SparseDriveBackbone=ResNet502026.03 | 8 | — | — | — | 0.61 | — | — | — | — | — | — | — | — | 2.55 | |
| DiffusionDriveBackbone=ResNet502026.03 | 8 | — | — | — | 0.57 | — | — | — | — | — | — | — | — | 1.35 | |
| SparseDriveType=Planning model, Backbone=ResNet-502026.05 | 8 | 0.29 | 0.58 | 0.96 | 0.61 | 1 | 5 | 0.18 | — | — | — | — | — | — | |
| SparseDrive-HATStage=2nd2025.12 | 8.4 | — | — | — | — | — | — | — | 0.6 | — | — | — | — | — | |
| MomADInput=Camera, Backbone=ResNet50, FPS=7.8, Hardware=RTX40902026.03 | 9 | 0.31 | 0.57 | 0.91 | 0.6 | 1 | 5 | 22 | — | 0.3 | 0.53 | 0.78 | 0.54 | — | |
| MomADBackbone=ResNet502026.03 | 9 | — | — | — | 0.6 | — | — | — | — | — | — | — | — | — | |
| FocalADBackbone=ResNet502026.03 | 9 | — | — | — | 0.6 | — | — | — | — | — | — | — | — | — | |
| BridgeADType=Planning model, Backbone=ResNet-502026.05 | 9 | 0.29 | 0.57 | 0.92 | 0.59 | 1 | 5 | 0.22 | — | — | — | — | — | — | |
| MomADType=Planning model, Backbone=ResNet-502026.05 | 9 | 0.31 | 0.57 | 0.91 | 0.6 | 1 | 5 | 0.22 | — | — | — | — | — | — | |
| DiffusionDrivereproduced=true2025.06 | 9 | 0.29 | 0.58 | 0.96 | 0.61 | 2 | 5 | 0.22 | — | — | — | — | — | — | |
| FocalAD2025.06 | 9 | 0.27 | 0.57 | 0.96 | 0.6 | 0 | 4 | 0.24 | — | — | — | — | — | — | |
| SparseDrivereproduced=true2025.06 | 10 | 0.3 | 0.58 | 0.96 | 0.61 | 1 | 5 | 0.23 | — | — | — | — | — | — | |
| DrivePIVLM-based=true, Ego Status=true2025.12 | 11 | 0.19 | 0.36 | 0.64 | 0.4 | 0 | 5 | 28 | — | — | — | — | — | — | |
| DriveTransformerBackbone=ResNet50, ego status=true2026.03 | 11 | — | — | — | 0.4 | — | — | — | — | — | — | — | — | — | |
| UniADBackbone=ResNet1012026.03 | 12 | — | — | — | 0.69 | — | — | — | — | — | — | — | — | 2.69 | |
| UniADType=Planning model, Backbone=ResNet-1012026.05 | 12 | 0.44 | 0.67 | 0.96 | 0.69 | 4 | 8 | 0.23 | — | — | — | — | — | — | |
| SparseDrive†Stage=2nd2025.12 | 12.3 | — | — | — | — | — | — | — | 0.63 | — | — | — | — | — | |
| SSRTraining Approach=SS-L, Ego status=true2025.12 | 13 | 0.19 | 0.36 | 0.62 | 0.39 | 0 | 10 | 20 | — | — | — | — | — | — | |
| PARA-DriveUsing Ego status=false2025.01 | 14 | 0.38 | 0.97 | 1.88 | 1.08 | — | — | — | — | — | — | — | — | — | |
| VADVLM-based=false, Ego Status=true2025.12 | 14 | 0.17 | 0.34 | 0.6 | 0.37 | 7 | 10 | 24 | — | — | — | — | — | — | |
| TOKENUsing Ego status=false2025.01 | 15 | 0.26 | 0.7 | 1.46 | 0.81 | — | — | — | — | — | — | — | — | — | |
| UniADUsing Ego status=false2025.01 | 15 | 0.47 | 1.09 | 1.92 | 1.16 | — | — | — | — | — | — | — | — | — | |
| WorldRFT (without RFT)Training Approach=SS-L, Ego status=false2025.12 | 15 | 0.21 | 0.44 | 0.76 | 0.47 | 10 | 11 | 23 | — | — | — | — | — | — | |
| DriveDreamerBackbone=-, ego status=true2026.03 | 15 | — | — | — | 0.29 | — | — | — | — | — | — | — | — | — | |
| World4DriveTraining Approach=SS-L, Ego status=false2025.12 | 16 | 0.23 | 0.47 | 0.81 | 0.5 | 2 | 12 | 33 | — | — | — | — | — | — | |
| GenADBackbone=ResNet502026.03 | 16 | — | — | — | 0.58 | — | — | — | — | — | — | — | — | 1.78 | |
| World4DriveType=Planning with world model, Backbone=ResNet-502026.05 | 16 | 0.23 | 0.47 | 0.81 | 0.5 | 2 | 12 | 0.33 | — | — | — | — | — | — | |
| PARA-DriveUsing Ego status=false2025.01 | 17 | 0.26 | 0.59 | 1.12 | 0.66 | — | — | — | — | — | — | — | — | — | |
| UncADTraining Approach=P-IL, Ego status=false2025.12 | 17 | 0.33 | 0.59 | 0.94 | 0.62 | 10 | 14 | 28 | — | — | — | — | — | — | |
| ForeSightType=Planning with world model, Backbone=ResNet-502026.05 | 18 | 0.36 | 0.55 | 0.93 | 0.62 | 4 | 12 | 0.37 | — | — | — | — | — | — | |
| PPADTraining Approach=P-IL, Ego status=false2025.12 | 19 | 0.31 | 0.56 | 0.87 | 0.58 | 8 | 12 | 38 | — | — | — | — | — | — | |
| GenADTraining Approach=P-IL, Ego status=false2025.12 | 19 | 0.28 | 0.49 | 0.78 | 0.52 | 8 | 14 | 34 | — | — | — | — | — | — | |
| LAW (Perception-based)Training Approach=P-IL, Ego status=false2025.12 | 19 | 0.24 | 0.46 | 0.76 | 0.49 | 8 | 10 | 39 | — | — | — | — | — | — | |
| GenADType=Planning model, Backbone=ResNet-502026.05 | 19 | 0.28 | 0.49 | 0.78 | 0.52 | 8 | 14 | 0.34 | — | — | — | — | — | — | |
| VAD+Reproduced=true2024.05 | 21 | 0.41 | 0.7 | 1.05 | 0.72 | 3 | 19 | 43 | — | — | — | — | — | — | |
| VADInput=Camera, Backbone=ResNet502026.03 | 21 | 0.41 | 0.7 | 1.05 | 0.72 | 3 | 19 | 43 | — | 0.36 | 0.66 | 0.91 | 0.64 | — | |
| VADBackbone=ResNet502026.03 | 21 | — | — | — | 0.72 | — | — | — | — | — | — | — | — | 3.07 | |
| VADVLM-based=false, Ego Status=false2025.12 | 22 | 0.41 | 0.7 | 1.05 | 0.72 | 7 | 17 | 41 | — | — | — | — | — | — | |
| VAD-BaseType=Planning model, Backbone=ResNet-502026.05 | 22 | 0.41 | 0.7 | 1.05 | 0.72 | 7 | 17 | 0.41 | — | — | — | — | — | — | |
| VADTraining Approach=P-IL, Ego status=false2025.12 | 23 | 0.41 | 0.7 | 1.05 | 0.72 | 7 | 18 | 43 | — | — | — | — | — | — | |
| OpenDriveVLA-7BVLM-based=true, Ego Status=true2025.12 | 25 | 0.2 | 0.58 | 1.21 | 0.66 | 0 | 22 | 55 | — | — | — | — | — | — | |
| PARA-DriveTraining Approach=P-IL, Ego status=false2025.12 | 25 | 0.25 | 0.46 | 0.74 | 0.48 | 14 | 23 | 39 | — | — | — | — | — | — | |
| PARA-DriveType=Planning model, Backbone=ResNet-502026.05 | 25 | 0.25 | 0.46 | 0.74 | 0.48 | 14 | 23 | 0.39 | — | — | — | — | — | — | |
| DriveWMBackbone=-2026.03 | 26 | — | — | — | 0.8 | — | — | — | — | — | — | — | — | — | |
| VADreproduced=true2025.06 | 26 | 0.41 | 0.7 | 1.05 | 0.72 | 11 | 24 | 0.42 | — | — | — | — | — | — | |
| UniAD2025.06 | 29 | 0.48 | 0.96 | 1.65 | 1.03 | 10 | 15 | 0.61 | — | — | — | — | — | — | |
| OmniDriveVLM-based=true, Ego Status=true2025.12 | 30 | 0.14 | 0.29 | 0.55 | 0.33 | 0 | 13 | 78 | — | — | — | — | — | — | |
| LAW (Perception-free)Training Approach=SS-L, Ego status=false2025.12 | 30 | 0.26 | 0.57 | 1.01 | 0.61 | 14 | 21 | 54 | — | — | — | — | — | — | |
| MomADInput=Camera, Backbone=ResNet50, Evaluation Protocol=UniAD, FPS=7.8, Hardware=RTX40902026.03 | 30 | 0.43 | 0.88 | 1.62 | 0.98 | 6 | 16 | 68 | — | 0.37 | 0.74 | 1.3 | 0.8 | — | |
| LAWBackbone=-2026.03 | 30 | — | — | — | 0.61 | — | — | — | — | — | — | — | — | — | |
| LAWType=Planning with world model, Backbone=ResNet-502026.05 | 30 | 0.26 | 0.57 | 1.01 | 0.61 | 14 | 21 | 0.54 | — | — | — | — | — | — | |
| UniADVLM-based=false, Ego Status=false2025.12 | 31 | 0.48 | 0.96 | 1.65 | 1.03 | 5 | 17 | 71 | — | — | — | — | — | — | |
| UniADTraining Approach=P-IL, Ego status=false2025.12 | 31 | 0.48 | 0.96 | 1.65 | 1.03 | 5 | 17 | 71 | — | — | — | — | — | — | |
| UniADInput=Camera, Backbone=ResNet101, Evaluation Protocol=UniAD, FPS=1.8, Hardware=A1002026.03 | 31 | 0.48 | 0.96 | 1.65 | 1.03 | 5 | 17 | 71 | — | 0.45 | 0.89 | 1.54 | 0.96 | — | |
| SparseDriveInput=Camera, Backbone=ResNet50, Evaluation Protocol=UniAD, Reproduction Status=Official Checkpoint, FPS=9.0, Hardware=RTX40902026.03 | 32 | 0.44 | 0.92 | 1.69 | 1.01 | 7 | 19 | 71 | — | 1.41 | 0.77 | — | 0.85 | — | |
| EOLiDAR-based=true2024.05 | 33 | 0.67 | 1.36 | 2.78 | 1.6 | 4 | 9 | 88 | — | — | — | — | — | — | |
| EOVLM-based=false, Ego Status=false2025.12 | 33 | 0.67 | 1.36 | 2.78 | 1.6 | 4 | 9 | 88 | — | — | — | — | — | — | |
| EponaTraining Approach=SS-L, Ego status=false2025.12 | 36 | 0.61 | 1.17 | 1.98 | 1.25 | 1 | 22 | 85 | — | — | — | — | — | — | |
| ORIONVLM-based=true, Ego Status=true2025.12 | 37 | 0.17 | 0.31 | 0.55 | 0.34 | 5 | 25 | 80 | — | — | — | — | — | — | |
| ORION (VLM)Backbone=-2026.03 | 37 | — | — | — | 0.63 | — | — | — | — | — | — | — | — | — | |
| DrivePIVLM-based=true, Ego Status=false2025.12 | 38 | 0.24 | 0.46 | 0.78 | 0.49 | 38 | 27 | 48 | — | — | — | — | — | — | |
| UniADUsing Ego status=false2025.01 | 40 | 0.48 | 0.89 | 1.47 | 0.95 | — | — | — | — | — | — | — | — | — | |
| FFLiDAR-based=true2024.05 | 43 | 0.55 | 1.2 | 2.54 | 1.43 | 6 | 17 | 107 | — | — | — | — | — | — | |
| FFVLM-based=false, Ego Status=false2025.12 | 43 | 0.55 | 1.2 | 2.54 | 1.43 | 6 | 17 | 107 | — | — | — | — | — | — | |
| VADInput=Camera, Backbone=ResNet50, Evaluation Protocol=UniAD2026.03 | 43 | 0.54 | 1.15 | 1.98 | 1.22 | 10 | 24 | 96 | — | 0.47 | 0.83 | 1.43 | 0.91 | — | |
| BEV-PlannerType=Planning model, Backbone=ResNet-502026.05 | 49 | 0.28 | 0.42 | 0.68 | 0.46 | 4 | 37 | 1.07 | — | — | — | — | — | — | |
| UniADReproduced=true2024.05 | 61 | 0.45 | 0.7 | 1.04 | 0.73 | 62 | 58 | 63 | — | — | — | — | — | — | |
| UniADInput=Camera, Backbone=ResNet101, FPS=1.8, Hardware=A1002026.03 | 61 | 0.45 | 0.7 | 1.04 | 0.73 | 62 | 58 | 63 | — | 0.41 | 0.68 | 0.97 | 0.68 | — | |
| ST-P32024.05 | 71 | 1.33 | 2.11 | 2.9 | 2.11 | 23 | 62 | 127 | — | — | — | — | — | — | |
| ST-P3VLM-based=false, Ego Status=false2025.12 | 71 | 1.33 | 2.11 | 2.9 | 2.11 | 23 | 62 | 127 | — | — | — | — | — | — | |
| ST-P3Training Approach=P-IL, Ego status=false2025.12 | 71 | 1.33 | 2.11 | 2.9 | 2.11 | 23 | 62 | 127 | — | — | — | — | — | — | |
| ST-P32025.06 | 71 | 1.33 | 2.11 | 2.9 | 2.11 | 23 | 62 | 1.27 | — | — | — | — | — | — | |
| OccNetTraining Approach=P-IL, Ego status=false2025.12 | 72 | 1.29 | 2.13 | 2.99 | 2.13 | 21 | 59 | 137 | — | — | — | — | — | — | |
| AutoVLAMethod Category=VLMs / VLAs with Reasoning2026.05 | — | 0.25 | 0.46 | 0.73 | 0.48 | — | — | — | — | — | — | — | — | — | |
| BEV-PlannerMethod Category=Training-based Policy2026.05 | — | 0.16 | 0.32 | 0.57 | 0.35 | — | — | — | — | — | — | — | — | — | |
| DriveVLMMethod Category=VLMs / VLAs with Reasoning2026.05 | — | 0.18 | 0.34 | 0.68 | 0.4 | — | — | — | — | — | — | — | — | — |