Reproduce “Does the Action Matter? Probing Action-Conditioned Semantic Forecasting as a World Model for 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
Binay Kumar Singh, Niels Da Vitoria Lobo
The authors test whether driving forecasters use actions when predicting semantic maps. They hold history and past actions fixed, replace the future action at test time, and score against recorded targets. On Cityscapes, FiLM gains 2.46 mIoU points over no action, but a shuffled action lowers its score by 23.17 points, below no action; this measures dependence, not counterfactual correctness.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.