Plug-In Diffusion Model for Sequential Recommendation (AAAI 2024)
Overview
This entry summarizes the AAAI 2024 paper "Plug-In Diffusion Model for Sequential Recommendation", published in the AAAI proceedings.
- Venue: AAAI Conference on Artificial Intelligence (AAAI 2024)
- Source: https://ojs.aaai.org/index.php/AAAI/article/view/28736
- Topic area: Sequential Recommendation
- Explores applying diffusion models to sequential recommendation, where the goal is to predict a user's next interactions from their behavior history.
- Positions the diffusion model as a plug-in module that complements existing sequential recommendation architectures, rather than a full replacement of established sequence encoders.
- Motivated by the gap between continuous diffusion-based generation and the discrete, large item-ID space typical of recommendation systems.
- BERT4Rec: Sequential Recommendation with Bidirectional Encoder Representations from Transformers
- 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
Key Points
Context in Sequential Recommendation Research
Sequential recommendation has evolved from matrix factorization through deep CTR models to Transformer-based encoders such as BERT4Rec, and more recently toward generative approaches. Diffusion models bring a probabilistic generative framework that can capture the uncertainty and diversity of users' future behavior, but adapting them to recommendation requires care around discrete item representations and inference cost.
Related Entries
Notes
Detailed quantitative results, dataset configurations, and ablation studies should be read from the original paper PDF at the AAAI proceedings page. This entry is based on the paper's title, venue, and publicly available metadata.