Reproduce “Scaling Action-Free Video Latent Prediction for Transferable Representations 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
Xingtai Gui, Xubin Zhou, Yucheng Zhou, Dongqian Guo, Jiahao Gong, Feiyang Tan, Jianbing Shen
ScaleDrive learns visual features by predicting masked regions in unlabeled driving video, including entire missing time slices, then freezes its encoder while planners learn trajectories. With architecture, training configuration, and budget fixed, NAVSIMv1 PDMS rises from 92.77 at 0.5M clips to 93.61 at 6M, versus 92.40 for V-JEPA2. This supports gains over the tested range, not real-world safety or indefinite scaling.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.