Reproduce “Synthesis Without Training: An Inference-Only Pipeline for Tabular, Temporal, and Relational Synthetic Data” 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
Zilong Zhao, Abdul Raheem, Jiayu Li, Sohei Arisaka, Darius Lim Hong Yi, Milad Abdollahzadeh, Uzair Javaid, Biplab Sikdar
GENSCRIPT replaces per-dataset training of synthetic-data generators with an inference-only pipeline. It profiles data, a frozen language model infers relationships, and a coding agent compiles an auditable sampler with explicit predicates. On Adult, it generated 32,561 rows with 0.000% violations of the education-to-education-num mapping. However, pairwise-correlation fidelity was lowest on four benchmarks, scoring 0.748 to 0.897; the constraint result is dataset-specific.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.