AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning

Zewei Zhou (University of California, Los Angeles) · Tianhui Cai (Columbia University) · Seth Zhao (UCLA Computer Science Department, University of California, Los Angeles) · Yun Zhang (Hong Kong University of Science and Technology) · Zhiyu Huang (University of California, Los Angeles) · Bolei Zhou (UCLA) · Jiaqi Ma (University of Illinois Urbana-Champaign)

autonomous driving autoregressive generationdiscrete actionsdual thinking modesfast thinkinggroup relative policy optimization planning performance qualitative results real-world datasets reinforcement fine-tuning semantic reasoning simulated datasetsslow thinkingsupervised fine-tuning trajectory planning vision-language-action

Recent advancements in Vision-Language-Action (VLA) models have shown promise for end-to-end autonomous driving by leveraging world knowledge and reasoning capabilities. However, current VLA models often struggle with physically infeasible action outputs, complex model structures, or unnecessarily long reasoning. In this paper, we propose AutoVLA, a novel VLA model that unifies reasoning and action generation within a single autoregressive generation model for end-to-end autonomous driving. AutoVLA performs semantic reasoning and trajectory planning directly from raw visual inputs and language instructions. We tokenize continuous trajectories into discrete, feasible actions, enabling direct integration into the language model. For training, we employ supervised fine-tuning to equip the model with dual thinking modes: fast thinking (trajectory-only) and slow thinking (enhanced with chain-of-thought reasoning). To further enhance planning performance and efficiency, we introduce a reinforcement fine-tuning method based on Group Relative Policy Optimization (GRPO), reducing unnecessary reasoning in straightforward scenarios. Extensive experiments across real-world and simulated datasets and benchmarks, including nuPlan, nuScenes, Waymo, and CARLA, demonstrate the competitive performance of AutoVLA in both open-loop and closed-loop settings. Qualitative results showcase the adaptive reasoning and accurate planning capabilities of AutoVLA in diverse scenarios.