Reproduce “DriveFoundry: Co-Evolving Data and Model for Efficient Closed-Loop Iteration in 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
Ziyang Yao, Yuncheng Jiang, Fang Li, Haochen Liu, Shaoqing Xu, Wen Zhang, Xiangyang Chu, Zibin Guo and 19 more
DriveFoundry links structured scene, candidate, and outcome records with a driving model that allocates additional reasoning according to estimated benefit and cost. Its shared model supports five tasks, while offline failure review updates records and training. In a tested 0.8B PhysicalAI/AlpaSim ablation, adaptive inference used 28% of full reasoning’s token consumption, with similar minADE@6; this is not a general guarantee.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.