Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Judge, Then Drive: A Critic-Centric Vision Language Action Framework for Autonomous Driving

About

Recent advances in vision language action (VLA) models have shown remarkable potential for autonomous driving by directly mapping multimodal inputs to control signals. However, previous VLA-based methods have not explicitly exploited the critic capability of VLAs to refine driving decisions, even though such capability has been well demonstrated in other LLM-based domains, thereby limiting their performance in complex closed-loop scenarios. In this work, we present a theoretically inspired two-stage framework, CriticVLA, which extends the role of VLAs from acting to judging. CriticVLA first generates a rough trajectory and then refines it through multimodal evaluation and single-step optimization guided by a VLA-based critic, yielding higher-quality driving behaviors. To support this process, we construct a large-scale synthetic dataset of 12.9 million annotated trajectories covering diverse driving scenarios, which enhances the critic's reasoning and refinement abilities. Extensive closed-loop experiments on the Bench2Drive benchmark show that CriticVLA significantly surpasses state-of-the-art baselines, achieving a 73.33% total success rate and delivering about 30% improvement in challenging scenarios.

Lijin Yang, Jianing Huang, Zhongzhan Huang, Shu Liu, Hao Yang• 2026

Related benchmarks

TaskDatasetResultRank
Closed-loop Autonomous DrivingBench2Drive
Driving Score (DS)88.02
74
Autonomous DrivingBench2Drive closed-loop (test)
SR (%)73.33
28
Closed-loop Autonomous DrivingBench2Drive v1 (test)
Success Rate (SR)73.33
19
Autonomous DrivingLongest6 v2
DS Score34
5
Inference EfficiencyLongest6 v2
Speed Ratio0.0146
4
Showing 5 of 5 rows

Other info

Follow for update