Reproduce “Structured Representation Probing for Risk Anticipation in End-to-End Autonomous Driving” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jiawei Liang, Tianrui Lou, Le Yang, Xianghao Jiao, Wei Wang, Xiaochun Cao
SRP monitors frozen driving policies for hard failures by reading their actor, ego, and map representations, without changing the policy or adding a perception-and-planning stack. It learns source-specific risk directions and pools element scores into a generic warning. On Bench2Drive with DriveTransformer, it reaches 0.833 ROC-AUC versus 0.715 for Temporal Consistency; simulated results do not establish explanations or safety.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.