Reproduce “LLMDiRec: LLM-Enhanced Intent Diffusion for Sequential 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
Bo-Chian Chen, Manel Slokom
LLMDiRec introduces a framework for sequential recommendation that integrates Large Language Model (LLM) semantics into an intent-aware diffusion model. This approach generates semantically coherent item augmentations and achieves state-of-the-art recommendation accuracy, particularly demonstrating up to a 160% improvement for long-tail items and 10% for cold-start users across various datasets.
Key insights
- Dynamic gated fusion effectively balances collaborative and semantic signals, outperforming static fusion methods in diverse data sparsity scenarios.
- Integrating LLM-based semantics into intent discovery and diffusion denoising leads to more discriminative and semantically coherent user intent representations (307% silhouette score improvement).
- Explicitly aligning collaborative and semantic embeddings via an alignment loss facilitates knowledge transfer and enhances the integration of dual item views.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.