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
Xiaomi EV
DriveFoundry links scene evidence, candidate actions, and outcome assessments in a five-task model with adaptive inference. It starts with Action and Value, requesting added task execution or latent tokens when expected benefit justifies the cost. On a curated held-out set, its 9B model records 0.83 m five-second ADE@1 versus 1.51 m for 34B Alpamayo 2 Super in an open-loop comparison.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.