[ICLR 2026] ReCogDrive: A Reinforced Cognitive Framework for End-to-End Autonomous Driving
Official Code Release of DiffSemanticFusion [including Mapless QCNet], which achieves SOTA in both nuScenes and NAVSIM
🚗 AUGUstus: Autonomous Driving VLA Model with Unified Generation and Understanding. 🏋️ 512 H800 GPUs · ~2 days · 318K+ video hours · 513B+ training tokens. ⚡ 12 Hz on-vehicle deployment. SFT + GRPO + NAVSIM.
A framework for unified end-to-end autonomous driving datasets processing
🔥 VITAL: From Visual Motion to Precise Action
[NeurIPS 2025] AGI-Elo: How Far Are We From Mastering A Task?
DriveTTO: Understanding Search and Selection in Test-Time Trajectory Optimization
Official implementation of "VLM-Guided Diffusion Models for Vehicle Trajectory Prediction" (World Wide Web, under review). Built on NAVSIM.
Geometric Safety Verdicts Need Not Be Learned for Driving World-Action Models🚀🚀🚀
Action-conditioned JEPA world models and latent planning. Main deployments for NAVSIM v1/v2 and nuScenes.
NAVSIM ablation: a 10-agent ability ladder with paired bootstrap CIs. What actually caps open-loop planning: kinematics, perception, or the map?
Ego trajectory prediction on NAVSIM from front-camera frames and kinematics with a small transformer
To associate your repository with the navsim topic, visit your repo's landing page and select "manage topics."