Reproduce “Attribute-Aware Controlled Product Generation with LLMs for E-commerce” 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
Virginia Negri, Víctor Martínez Gómez, Sergio A. Balanya, Subburam Rajaram
The authors present a method for generating labeled product listings from existing examples by changing one attribute while making text agree with it, conflict subtly, or omit it. The method aims to preserve listing coherence and marketplace constraints. In a FLAN-T5-base test on MAVE, synthetic-only training reached 60.48% accuracy, versus 60.79% for original-only training; this result covers only correct examples.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.