Reproduce “Neural Symbolic Regression that Scales” 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
Luca Biggio, Tommaso Bendinelli, Alexander Neitz, Aurelien Lucchi, Giambattista Parascandolo
NeSymReS infers symbolic equations from input-output examples by pre-training on equation skeletons, then proposing skeletons for new measurements and fitting constants. Increasing pre-training data from 10K to 10M skeletons improved accuracy on five evaluated datasets. On AI-Feynman, it reached Genetic Programming’s maximum accuracy in more than three orders of magnitude less test time, but its skeletons were included in pre-training.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.