NoRD: A Data-Efficient Vision-Language-Action Model that Drives without Reasoning

Researchers introduced NoRD (No Reasoning for Driving), a Vision-Language-Action model for autonomous driving that achieves competitive performance using 60% less training data and no reasoning annotations. The model incorporates Dr. GRPO algorithm to overcome difficulty bias issues in reinforcement learning, demonstrating successful results on Waymo and NAVSIM benchmarks.

• →NoRD achieves competitive autonomous driving performance with 3x fewer tokens and 60% less training data than existing VLA models.
• →The model eliminates the need for expensive reasoning annotations while maintaining performance on industry benchmarks.
• →Standard Group Relative Policy Optimization (GRPO) fails on small, reasoning-free datasets due to difficulty bias.
• →Dr. GRPO algorithm successfully mitigates difficulty bias from high-variance scenarios in autonomous driving training.
• →The approach enables more efficient autonomous driving systems by reducing data collection and annotation costs.