NAVSIMv-1

NAVSIMv-1 is a dataset/benchmark that contains driving scenarios used to evaluate the performance of Vision-Language-Action models in terms of their spatial intelligence and ability to follow language-guided driving policies.

Papers using NAVSIMv-1 (5)

DriveStack-VLA: Render-Teacher Alignment for BEV-Based DeepStack Vision-Language-Action Model2026LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Driving2026DriveWorld-VLA: Unified Latent-Space World Modeling with Vision-Language-Action for Autonomous Driving2026DriveFine: Refining-Augmented Masked Diffusion VLA for Precise and Robust Driving2026MindDrive: An All-in-One Framework Bridging World Models and Vision-Language Model for End-to-End Autonomous Driving2025