Reproduce “Multimodal Masked Diffusion Autoencoder for Modeling Impaired 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

John Gideon, Anshul Gupta, Megan Applegate-Kenton, Kayli Battel, Todd J Rowell, Patricio Reyes Gomez, Laporsha Dees, Emily Sumner and 2 more

ID-MAE trains a driver-state model to predict masked impairment labels while reconstructing hidden gaze and vehicle trajectories from multimodal driving data. It uses a shared masked-diffusion model, with video as context, so behavior provides a learning signal alongside labels. On DIVERT, its location-matched impairment-detection P-AUC is 0.706, versus 0.590 without reconstruction loss; this is simulator evidence, not public-road validation.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.