Reproduce “LoopGR: Parameter-Efficient Scaling for Generative Recommendation via Recurrent Depth” 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

Huizhe Zhang, Yudong Li, Liang Chen, Chengxiang Zhuo, Zang Li

LoopGR generates next-item predictions as semantic-ID sequences from interaction histories. It shares a decoder across recurrent passes and uses an updater to revise history context from emerging predictions, separating computational depth from parameter growth. At approximately 0.5 GFLOPs, it uses 1.3M parameters versus 2.7M for an encoder–decoder, and reports higher NDCG@5 across compared computation levels; results are offline.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.