Reproduce “Taxonomy-Guided Zero-Shot Recommendations with LLMs” 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

Yueqing Liang, Liangwei Yang, Chen Wang, Xiongxiao Xu, Philip S. Yu, Kai Shu

The TAXREC framework enhances zero-shot recommendation systems by guiding Large Language Models with a pre-generated, domain-specific taxonomy. This approach improves recommendation quality and efficiency by structuring item information and constraining LLM output, significantly outperforming traditional and direct LLM methods on Movie and Book datasets.

Key insights

  • Integrating a structured taxonomy significantly improves LLMs' ability to act as effective zero-shot recommenders by providing necessary clarity and constraint.
  • The pre-processing step of item categorization effectively addresses prompt length limitations and enriches vague item information, making LLMs viable for systems with large item pools.
  • Guiding LLM generation to specific, taxonomy-defined features ensures that recommendations are actionable, can be accurately matched to actual candidate items, and prevents out-of-domain suggestions.

Summary by alphaXiv.

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