List your company

Autonomous Driving Planning on NAVSIM Navhard v2 (Stage 2)

0.881NCPDM-Closed
0.743720.779360.8150.85064Feb 5, 2026

Evaluation results

MethodLinks
PDM-ClosedPrivileged=true2026.020.8810.9060.9630.98510.915—
Constrained GRPO, ScAdvReward Formulation=Scalarized Advantages (ScAdv)2026.020.8530.830.8970.9860.8630.9670.516
Constrained GRPO, ScRewReward Formulation=Scalarized Rewards (ScRew)2026.020.810.7840.8660.9760.9390.9520.436
SFT Checkpoint2026.020.80.7560.8690.9810.8780.9630.444
GRPO, ScAdvReward Formulation=Scalarized Advantages (ScAdv), Weights=Informed Weights2026.020.7780.7310.8310.9720.9440.9620.385
GRPO, ScAdvReward Formulation=Scalarized Advantages (ScAdv), Weights=Equal Weights2026.020.7670.7050.8130.9710.9810.9490.37
GRPO, ScRewReward Formulation=Scalarized Rewards (ScRew), Weights=Informed Weights2026.020.7540.6950.7910.9690.9860.9540.35
GRPO, ScRewReward Formulation=Scalarized Rewards (ScRew), Weights=Equal Weights2026.020.7510.6960.7870.9680.9860.9540.347
GRPO with EPDMSReward Configuration=Direct EPDMS optimization2026.020.7490.7590.8290.9690.9790.9450.381