LISRec: Modeling User Preferences with Learned Item Shortcuts for Sequential Recommendation

2025/05/28 by Haidong Xin, Xin, Haidong, Zhenghao Liu +16 · 1 citation
Computer Science · Decision Sciences · #Advanced Bandit Algorithms Research #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Mobile Crowdsensing and Crowdsourcing #Recommender Systems and Techniques

paper · pdf · doi:10.48550/arxiv.2505.22130

openalex publication_date 2025/05/28 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/03

Abstract

User-item interaction histories are pivotal for sequential recommendation systems but often include noise, such as unintended clicks or actions that fail to reflect genuine user preferences. To address this, we propose Learned Item Shortcuts for Sequential Recommendation (LISRec), a novel framework that explicitly captures stable preferences by extracting personalized semantic shortcuts from historical interactions. LISRec first learns task-agnostic semantic representations to assess item similarities, then constructs a personalized semantic graph over all user-interacted items. By identifying the maximal semantic connectivity subset within this graph, LISRec selects the most representative items as semantic shortcuts to guide user preference modeling. This focused representation filters out irrelevant actions while preserving the diversity of genuine interests. Experimental results on the Yelp and Amazon Product datasets illustrate that LISRec achieves a 13% improvement over baseline recommendation models, showing its effectiveness in capturing stable user interests. Further analysis indicates that shortcut-based histories better capture user preferences, making more accurate and relevant recommendations. All codes and datasets are available at https://github.com/NEUIR/LISRec.

Citations

• Personalized Denoising Implicit Feedback for Robust Recommender System
• HLLM: Enhancing Sequential Recommendations via Hierarchical Large Language Models for Item and User Modeling
• Double Correction Framework for Denoising Recommendation
• SSDRec: Self-Augmented Sequence Denoising for Sequential Recommendation
• TinyLlama: An Open-Source Small Language Model
• Collaborative Word-based Pre-trained Item Representation for Transferable Recommendation
• Text Matching Improves Sequential Recommendation by Reducing Popularity Biases
• Frequency Enhanced Hybrid Attention Network for Sequential Recommendation
• Where to Go Next for Recommender Systems? ID- vs. Modality-based Recommender Models Revisited
• LLaMA: Open and Efficient Foundation Language Models
• LightGCL: Simple Yet Effective Graph Contrastive Learning for Recommendation
• Denoising and Prompt-Tuning for Multi-Behavior Recommendation
• Towards Reasoning in Large Language Models: A Survey
• Decoupled Side Information Fusion for Sequential Recommendation
• Zero-Shot Recommender Systems
• Dynamic Graph Neural Networks for Sequential Recommendation
• Non-invasive Self-attention for Side Information Fusion in Sequential Recommendation
• OpenMatch: An Open Source Library for Neu-IR Research
• Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
• CosRec: 2D Convolutional Neural Networks for Sequential Recommendation
• Warm Up Cold-start Advertisements: Improving CTR Predictions via Learning to Learn ID Embeddings
• BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from Transformer
• Self-Attentive Sequential Recommendation
• A Simple Convolutional Generative Network for Next Item Recommendation
• Personalized Top-N Sequential Recommendation via Convolutional Sequence Embedding
• Fusing Similarity Models with Markov Chains for Sparse Sequential Recommendation
• Context-aware Sequential Recommendation
• Session-based Recommendations with Recurrent Neural Networks
• Image-based Recommendations on Styles and Substitutes

Cited by

• LLMsPark: A Benchmark for Evaluating Large Language Models in Strategic Gaming Contexts

Related

• Text Matching Improves Sequential Recommendation by Reducing Popularity Biases
• ULMRec: User-centric Large Language Model for Sequential Recommendation
• Reformulating Sequential Recommendation: Learning Dynamic User Interest with Content-enriched Language Modeling
• CESRec: Constructing Pseudo Interactions for Sequential Recommendation via Conversational Feedback
• Next Item Recommendation with Self-Attention