Reinforcing End-to-End Autonomous Driving Safety via Negative-Aware Planning in a Unified World-Action Model
Reinforcing End-to-End Autonomous Driving Safety via Negative-Aware Planning in a Unified World-Action Model
This research presents NegWAM, a safety post-training framework for unified world-action models (WAMs) that learns explicit safety boundaries from near-expert failure trajectories.
The paper introduces NegWAM, a post-training framework for world-action models that incorporates negative trajectories to improve safety in end-to-end autonomous driving, reporting specific performance gains on NAVSIM v2 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
Reported numbers
EPDMS score on navhard
40.4
View original evidence
trajectory compatibility. During training, NegWAM jointly optimizes the flow-matching and near-expert negative objectives, while the auxiliary rider branch is discarded at inference, incurring no additional overhead. On NAVSIM v2, NegWAM achieves 40.4 EPDMS on navhard, a 4.1% relative improvement over the strongest compared scorer-free WAM baseline, and achieves 93.3 EPDMS on navtest. These results demonstrate that N
Open source S4EPDMS score on navtest
93.3
View original evidence
trajectory compatibility. During training, NegWAM jointly optimizes the flow-matching and near-expert negative objectives, while the auxiliary rider branch is discarded at inference, incurring no additional overhead. On NAVSIM v2, NegWAM achieves 40.4 EPDMS on navhard, a 4.1% relative improvement over the strongest compared scorer-free WAM baseline, and achieves 93.3 EPDMS on navtest. These results demonstrate that N
Open source S4
- code: Available Artifact link Open source S5
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2611.08111v1 Announce Type: new Abstract: World-action models (WAMs) jointly model future traffic states and driving actions for end-to-end autonomous driving in intelligent transportation systems. However, their action learning remains dominated by expert-trajectory imitation, leaving unsafe alternatives close to expert trajectories insufficiently constrained. To improve WAM robustness in safety-critical scena
Open source S2
trajectory compatibility. During training, NegWAM jointly optimizes the flow-matching and near-expert negative objectives, while the auxiliary rider branch is discarded at inference, incurring no additional overhead. On NAVSIM v2, NegWAM achieves 40.4 EPDMS on navhard, a 4.1% relative improvement over the strongest compared scorer-free WAM baseline, and achieves 93.3 EPDMS on navtest. These results demonstrate that N
Open source S4
egWAM achieves strong performance in safety-critical scenarios while preserving routine-driving performance. Zero-shot closed-loop evaluation on HUGSIM further demonstrates cross-benchmark generalization. The code is available at https://github.com/Zizizi-hao/NegWAM.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org