Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLMDiRec: LLM-Enhanced Intent Diffusion for Sequential Recommendation

Bo-Chian Chen, Manel Slokom

TL;DR

LLMDiRec tackles the semantic grounding gap in sequential recommendations by embedding LLM-derived semantic representations into an intent-aware diffusion framework. It introduces a dual-view item representation with gated fusion, an intent-guided diffusion process using semantic prototypes, and a multi-task objective that jointly optimizes next-item prediction, denoising, contrastive learning, and embedding alignment. Across five public datasets, LLMDiRec consistently surpasses state-of-the-art baselines, with particularly large gains on long-tail items and cold-start users, demonstrating the value of semantic guidance for intent modeling. The approach yields more interpretable, semantically coherent user intents and shows robust performance across domains, suggesting practical impact for sparse data regimes and real-world recommender systems.

Abstract

Existing sequential recommendation models, even advanced diffusion-based approaches, often struggle to capture the rich semantic intent underlying user behavior, especially for new users or long-tail items. This limitation stems from their reliance on ID-based embeddings, which lack semantic grounding. We introduce LLMDiRec, a new approach that addresses this gap by integrating Large Language Models (LLMs) into an intent-aware diffusion model. Our approach combines collaborative signals from ID embeddings with rich semantic representations from LLMs, using a dynamic fusion mechanism and a multi-task objective to align both views. We run extensive experiments on five public datasets. We run extensive experiments on five public datasets. We demonstrate that \modelname outperforms state-of-the-art algorithms, with particularly strong improvements in capturing complex user intents and enhancing recommendation performance for long-tail items.

LLMDiRec: LLM-Enhanced Intent Diffusion for Sequential Recommendation

TL;DR

LLMDiRec tackles the semantic grounding gap in sequential recommendations by embedding LLM-derived semantic representations into an intent-aware diffusion framework. It introduces a dual-view item representation with gated fusion, an intent-guided diffusion process using semantic prototypes, and a multi-task objective that jointly optimizes next-item prediction, denoising, contrastive learning, and embedding alignment. Across five public datasets, LLMDiRec consistently surpasses state-of-the-art baselines, with particularly large gains on long-tail items and cold-start users, demonstrating the value of semantic guidance for intent modeling. The approach yields more interpretable, semantically coherent user intents and shows robust performance across domains, suggesting practical impact for sparse data regimes and real-world recommender systems.

Abstract

Existing sequential recommendation models, even advanced diffusion-based approaches, often struggle to capture the rich semantic intent underlying user behavior, especially for new users or long-tail items. This limitation stems from their reliance on ID-based embeddings, which lack semantic grounding. We introduce LLMDiRec, a new approach that addresses this gap by integrating Large Language Models (LLMs) into an intent-aware diffusion model. Our approach combines collaborative signals from ID embeddings with rich semantic representations from LLMs, using a dynamic fusion mechanism and a multi-task objective to align both views. We run extensive experiments on five public datasets. We run extensive experiments on five public datasets. We demonstrate that \modelname outperforms state-of-the-art algorithms, with particularly strong improvements in capturing complex user intents and enhancing recommendation performance for long-tail items.
Paper Structure (22 sections, 1 equation, 3 figures, 4 tables)

This paper contains 22 sections, 1 equation, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Sequential Recommendation with Diffusion Models
  4. LLM Integration in Recommendation
  5. Intent Modeling in Sequential Recommendation
  6. Our LLMDiRec Recommender
  7. Dual-View Item Representation
  8. Intent-Aware Diffusion with Semantic Guidance
  9. Semantic-Enhanced Intent Discovery
  10. Conditional Denoising with Intent Signals
  11. Intent-Consistent Augmentation
  12. Multi-task Optimization
  13. Experimental Setup
  14. Datasets
  15. LLM Prompt Design
  16. ...and 7 more sections

Figures (3)

  • Figure 1: An illustration of semantically enhanced dual-view model and how random data augmentation (view 1, view 2) can disrupt semantic consistency. A user's interaction sequence contains two distinct intents: "For entertainment" and "For school." Augmentation methods like cropping or substitution can break the coherence of these intents, leading to flawed intent representations
  • Figure 2: Architecture of LLMDiRec. The model fuses collaborative and semantic embeddings, encodes the sequence, and uses intent clusters to guide a diffusion process for contrastive learning, all optimized under a multi-task objective.
  • Figure 3: t-SNE visualization of learned sequence representations on the Toys dataset for InDiRec (left) and LLMDiRec (right). LLM-enhanced representations form more distinct and coherent clusters.