Autonomous Driving on CARLA Leaderboard
2,237.31ReturnDrQv2-VMR
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| DrQv2-VMRBase RL=DrQv2, Regularization=Value-Margin Regularization2026.02 | 2,237.31 | 72 | 89 | 60 | 78 | 3.56 | 9 | 0.002 | 0 | 16 | 2 | |
| DrQv2-CLIPBase RL=DrQv2, Alignment=Conditional Contrastive Action Alignment2026.02 | 2,188.54 | 77 | 90 | 57 | 77 | 3.6 | 13 | 0.002 | 0 | 14 | 1 | |
| VisualModel Size=0.1B, Input Modality=BEV (192×192×3)2026.02 | 2,044.01 | 65 | 86 | 52 | 70 | 3.63 | 5 | 0.001 | 1 | 15 | 8 | |
| TD3Algorithm Type=Baseline RL2026.02 | 1,703.83 | 53 | 89 | 37 | 58 | 3.48 | 18 | 0.004 | 0 | 21 | 3 | |
| DrQv2-AWAGBase RL=DrQv2, Guidance=Advantage-Weighted Action Guidance2026.02 | 1,626.35 | 61 | 90 | 46 | 66 | 3.68 | 6 | 0.001 | 1 | 13 | 2 | |
| SAC-VMRBase RL=SAC, Regularization=Value-Margin Regularization2026.02 | 1,623.34 | 53 | 89 | 31 | 58 | 3.63 | 7 | 0.002 | 0 | 20 | 2 | |
| DrQv2Algorithm Type=Baseline RL2026.02 | 1,507.66 | 56 | 88 | 38 | 61 | 3.72 | 9 | 0.002 | 1 | 19 | 2 | |
| SAC-AWAGBase RL=SAC, Guidance=Advantage-Weighted Action Guidance2026.02 | 556.33 | 28 | 82 | 8 | 33 | 5.21 | 8 | 0.002 | 1 | 63 | 15 | |
| SACAlgorithm Type=Baseline RL2026.02 | 527.03 | 21 | 90 | 3 | 23 | 3.73 | 15 | 0.003 | 1 | 50 | 5 |