Reproduce “Efficient Learning of Mechanism Designs and Compositionality” 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
João Pedro Gandarela, Thiago Rios, Stefan Menzel, André Freitas
The authors represent linkages as Assur groups and directed solve graphs, making explicit which modules are solved together and in what order. They compare Assur-derived features with flattened adjacency and position inputs using the same three-layer MLP for curve prediction. Across training budgets of 50 to 10,000 mechanisms, the features show a 1.8× sample-efficiency advantage that narrows at larger budgets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.