Perceives but Doesn't Brake: A Perception-Action Binding Gap in Reasoning Driving VLAs

Key points

• We characterize this failure as a perception-action binding gap: hazard information is available inside the model but exerts insufficient control over trajectory generation.1
• Using matched simulated counterfactuals and logged driving replay, we show that in-path hazards are actually encoded in the visual representation, but longitudinal plans are dominated by ego-speed history and respond only weakly to hazard proximity or the model's own reasoning trace.1
• Further analysis shows that the trajectory generator is capable of braking but sometimes fails to do so due to the perception-action binding gap.1
• These results separate hazard perception from hazard control in driving VLAs, show that unsafe behavior can arise even when the model has both the evidence and the ability to act on it, and propose a simple fix for bridging this binding gap.1

Sources (1)

• [1]Perceives but Doesn't Brake: A Perception-Action Binding Gap in Reasoning Driving VLAs
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 01:21 PM

  Reasoning-augmented vision-language-action (VLA) driving models can perceive and reason about safety-critical hazards yet fail to translate that evidence into safe actions.

  We characterize this failure as a perception-action binding gap: hazard information is available inside the model but exerts insufficient control over trajectory generation.

Extractive summary: sentences quoted from the sources.