Reproduce “Emergent Models: Intelligence from Tiny Substrates” 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
Giacomo Bocchese, Nicola Giacobbo, Etienne Guichard, James Wiles, Akshaj Devireddy
Emergent Models frame computation as searching for an initial state whose repeated local updates produce task behavior. An encoder writes the input; after repeated updates, a halting rule ends computation and a decoder reads the answer. On x mod 4, EM43 achieved 100% exact-match accuracy on training inputs [0, 30] and extrapolation inputs [30, 1000], averaged over eight runs; this does not establish general arithmetic ability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.