Reproduce “SNIP++: Fine-Grained Symbolic-Numerical Alignment for Symbolic Regression” 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
Benjamin Léger, Shubham Gupta, Samy Mammeri, Kazem Meidani, Cem Subakan, Christian Gagné
SNIP++ trains symbolic-regression representations to align mathematical expressions with numerical behavior, including at the level of subexpressions and controlled edits. It adds tree-position information and supervises edited parts and full expressions alongside their behaviors. On held-out local-edit tests, correct matches receive higher average percentile ranks than with SNIP in both directions; this does not establish better end-to-end equation discovery.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.