XLNet4Rec: Transformer-Based Recommendations Using Users' Long-Term and Short-Term Interests (ICMLA 2023)

Forum topic · 小凯 · 2026-07-05

Summary

This forum post is a catalog entry for XLNet4Rec, a research paper presented at ICMLA 2023 that applies the Transformer architecture to recommendation by modeling both users' long-term and short-term interests. The entry links to the paper's IEEE Xplore record and places it in a Sequential Recommendation reading-list section alongside related works such as BERT4Rec, EAGER, Mamba4Rec, P5, LLMCDSR, and on-device session-based recommendation. The surrounding commentary situates sequential recommendation within the broader search, recommendation, and personalization landscape, discussing pipeline components like representation learning, retrieval, reranking, and generation, as well as practical engineering constraints such as latency budgets, data quality, evaluation trustworthiness, and safety. The post is largely a templated summary based on metadata; it does not reproduce the paper's detailed methods or quantitative results, so readers should consult the original IEEE publication for experimental specifics.

XLNet4Rec: Transformer-Based Recommendations Using Users' Long-Term and Short-Term Interests (ICMLA 2023)

This post is an index entry from a Sequential Recommendation reading list on zhichai.net. It catalogs the paper "XLNet4Rec: Recommendations Based on Users' Long-Term and Short-Term Interests Using Transformer", published at ICMLA 2023.

Key points

Paper: XLNet4Rec applies a Transformer-based approach to recommendation, explicitly modeling the combination of a user's long-term interests and short-term interests.
Venue: ICMLA 2023 (International Conference on Machine Learning and Applications); the paper is available via IEEE Xplore: https://ieeexplore.ieee.org/abstract/document/10460010
Category: Sequential Recommendation.

Context in the reading list

The entry is cross-referenced with related sequential-recommendation works, including:

BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations
EAGER: Two-Stream Generative Recommender with Behavior-Semantic Collaboration
Efficient On-Device Session-Based Recommendation (ACM TORS)
How to Index Item IDs for Recommendation Foundation Models (P5, SIGIR)
LLMCDSR: Enhancing Cross-Domain Sequential Recommendation with Large Language Models
Mamba4Rec: Towards Efficient Sequential Recommendation with Selective State Space Models

Editorial notes

The forum post is a templated summary built from metadata rather than a detailed review. Its broader commentary discusses general themes in modern search/recommendation systems — retrieval-rerank-generation pipelines, LLM-era system design, latency and cost constraints, evaluation reliability, and safety — but it does not reproduce the paper's specific architecture, training procedure, datasets, or experimental numbers. Readers interested in quantitative results should consult the original IEEE publication directly.

Related resources

Original paper: IEEE Xplore record
Companion entries in the same Sequential Recommendation section (see list above)

XLNet4Rec: Transformer-Based Recommendations Using Users' Long-Term and Short-Term Interests (ICMLA 2023)

Summary

XLNet4Rec: Transformer-Based Recommendations Using Users' Long-Term and Short-Term Interests (ICMLA 2023)

Key points

Context in the reading list

Editorial notes

Related resources

Suggested reading approach

Tags