STEER: Route-Aware Adaptive Reasoning for Autonomous Driving
STEER: Route-Aware Adaptive Reasoning for Autonomous Driving
This research presents STEER, a novel strategy to mitigate over-reasoning in vision-language action (VLA) models for autonomous driving. By combining uncertainty-aware rollout routing with cross-route advantage credit assignment, the method dynamically adjusts inference complexity without requiring subjective scene-difficulty annotations. Experiments on NAVSIM v1 and v2 show that STEER achieves state-of-the-art planning quality while improving the Pareto front of inference efficiency compared to prior adaptive reasoning methods.
The paper introduces STEER, a method for autonomous driving that optimizes the balance between chain-of-thought reasoning and direct responses without relying on labor-intensive scene-difficulty labels, demonstrating competitive planning quality and inference efficiency on NAVSIM benchmarks.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- Control
- Reported autonomousOpen source S2
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2611.00718v1 Announce Type: new Abstract: Vision-language action (VLA) models excel for end-to-end autonomous driving, yet their tendency to over-invoke chain-of-thought (CoT) reasoning mirrors a human cognitive pitfall: overthinking. Just as deliberate reasoning can slow or mislead human judgment in routine tasks, excessive CoT in VLA incurs unnecessary compute overhead and can paradoxically degrade planning p
Open source S2
ing to align routing diversity with model uncertainty; and (ii) cross-route advantage credit assignment, which introduces a differential metric to reinforce optimal routing decisions based on environmental rewards rather than subjective labels. Extensive experiments on NAVSIM v1/v2 demonstrate that STEER is competitive with state-of-the-art methods in both planning quality and the Pareto front of inference efficiency
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org