Learning from Mistakes: Post-Training for Driving VLA with Takeover Data
This addresses safety limitations in autonomous driving systems for real-world deployment, though it builds incrementally on existing post-training methods.
The paper tackles the problem of distribution shift in Vision-Language-Action (VLA) models for autonomous driving by proposing TakeVLA, a post-training framework that uses takeover data to improve safety margins and performance. It achieves state-of-the-art results with a 4.93 increase in driving score and an 11.76% boost in average time-to-collision (TTC) on the Bench2Drive benchmark.
Current Vision-Language-Action (VLA) paradigms in end-to-end autonomous driving rely on offline training from static datasets, leaving them vulnerable to distribution shift. Recent post-training methods use takeover data to mitigate this by augmenting the dataset with high-quality expert takeover samples, yet they suffer from two key limitations: supervision restricted to the period after the takeover moments leads to policies with limited safety margins, and passive preference optimization lacks active exploration for optimal performance. In this paper, we propose TakeVLA, a novel VLA post-training framework that overcomes these shortcomings through two complementary innovations. First, we introduce pre-takeover language supervision, which allows the VLA to learn from mistakes proactively. By explicitly teaching the model about what to do in error-prone situations, we cultivate a precautionary mindset that anticipates hazards early and substantially enlarges safety margins. Second, we propose Scenario Dreaming, a reinforcement fine-tuning paradigm that operates in reconstruceted takeover scenarios, encouraging active exploration beyond mere preference fitting. Experiments on the Bench2Drive benchmark demonstrate that TakeVLA achieves state-of-the-art closed-loop performance, surpassing the strong VLA baseline SimLingo by 4.93 in driving score, with an enhanced safety margin as evidenced by an 11.76% increase in average TTC.