Reproduce “Modeling Behavioral Intensity and Transitions for Generative Recommendation” 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

Wenxuan Yang, Xiaoyang Xu, Hanyu Zhang, Zhexuan Xu, Wanqiang Xiong, Zhaoqun Chen

BITRec is a generative recommendation model designed to improve multi-behavior sequential recommendations by explicitly modeling behavioral intensity heterogeneity and directional transition structures. The system consistently outperformed state-of-the-art baselines on four large-scale datasets, achieving significant gains such as 17.83% on HR@10 and 17.55% on NDCG@10 on the Taobao dataset.

Key insights

  • Dependencies in multi-behavior sequences are not uniformly present but should be selectively triggered based on behavioral semantics and the structural relationship between actions.
  • Explicitly differentiating behavioral intensity (exploration vs. commitment) through stratified attention and adaptive fusion is crucial for accurately discerning user intent and improving conversion prediction.
  • Incorporating explicit, directional transition structures into attention mechanisms, along with temporal dynamics and item-level consistency, enhances the model's understanding of user decision paths.

Summary by alphaXiv.

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