Current sequential recommender systems are proposed to tackle the dynamic user preference learning with various neural techniques, such as Transformer and Graph Neural Networks (GNNs). However, inference from highly sparse behavior data may hinder representation ability of pattern encoding. To address label shortage issue, contrastive (CL) methods recently perform augmentation in two fashions: (i) randomly corrupting sequence (e.g. stochastic masking, reordering); (ii) aligning...

Graph neural networks (GNNs) have emerged as the state-of-the-art paradigm for collaborative filtering (CF). To improve representation quality over limited labeled data, contrastive learning has attracted attention in recommendation and benefited graph-based CF model recently. However, success of most methods heavily relies on manually generating effective views heuristic-based data augmentation. This does not generalize across different datasets downstream tasks, which is difficult to be...

With the growth of high-dimensional sparse data in web-scale recommender systems, computational cost to learn high-order feature interaction CTR prediction task largely increases, which limits use models real industrial applications. Some recent knowledge distillation based methods transfer from complex teacher shallow student for accelerating online model inference. However, they suffer degradation accuracy process. It is challenging balance efficiency and effectiveness models. To address...

Click-Through Rate (CTR) prediction, which aims to estimate the probability that a user will click an item, is essential component of online advertising. Existing methods mainly attempt mine interests from users' historical behaviours, contain directly interacted items. Although these have made great progress, they are often limited by recommender system's direct exposure and inactive interactions, thus fail all potential interests. To tackle problems, we propose Neighbor-Interaction based...

GNN-based recommendation systems have been successful in capturing complex user-item interactions using multi-hop message passing. However, these methods often struggle to handle the dynamic nature of interactions, making it challenging adapt changes user preferences and new data distributions. This limits their scalability performance real-world scenarios. In our study, we propose a framework called GraphPro that combines graph pre-training with prompt learning an efficient way. unique...

A Session-Based Recommendation (SBR) seeks to predict users’ future item preferences by analyzing their interactions with previously clicked items. In recent approaches, Graph Neural Networks (GNNs) have been commonly applied capture relations within a session infer user intentions. However, these GNN-based methods typically struggle feature ambiguity between the sequential information and conversion an graph, which may impede model’s ability accurately this article, we propose novel...

Traditional Click-Through Rate (CTR) prediction models are usually trained and deployed in a single scenario. However, large-scale commercial platforms contain multiple recommendation scenarios, the traffic characteristics of which may be significantly different. Recent studies have proved that learning unified model to serve scenarios is effective improving overall performance. most existing approaches suffer from various limitations respectively, such as insufficient distinction modeling,...

Click-Through Rate (CTR) prediction on cold users is a challenging task in recommender systems. Recent researches have resorted to meta-learning tackle the cold-user challenge, which either perform few-shot user representation learning or adopt optimization-based meta-learning. However, existing methods suffer from information loss inefficient optimization process, and they fail explicitly model global preference knowledge crucial complement sparse insufficient of users. In this paper, we...

Click-through Rate (CTR) prediction in real-world recommender systems often deals with billions of user interactions every day. To improve the training efficiency, it is common to update CTR model incrementally using new incremental data and a subset historical data. However, feature embeddings get stale when corresponding features do not appear current In next period, would have performance degradation on samples containing features, which we call staleness problem. mitigate this problem,...

The rise of online multi-modal sharing platforms like TikTok and YouTube has enabled personalized recommender systems to incorporate multiple modalities (such as visual, textual, acoustic) into user representations. However, addressing the challenge data sparsity in these remains a key issue. To address this limitation, recent research introduced self-supervised learning techniques enhance systems. methods often rely on simplistic random augmentation or intuitive cross-view information,...

Modern recommender systems aim to deeply understand users' complex preferences through their past interactions. While deep collaborative filtering approaches using Graph Neural Networks (GNNs) excel at capturing user-item relationships, effectiveness is limited when handling sparse data or zero-shot scenarios, primarily due constraints in ID-based embedding functions. To address these challenges, we propose a model-agnostic recommendation instruction-tuning paradigm that seamlessly...

The construction of social credit system is an important prerequisite and objective need for the orderly healthy development market economy. Through mechanism information sharing joint reward punishment mechanism, it reduces asymmetry reveals merits demerits subject credit. It integrates whole forces to praise good faith, punish breach trust, creates a environment. common phenomenon that enterprises are faced with financing constraints which not conducive growth enterprises. This paper...

GNN-based recommenders have excelled in modeling intricate user-item interactions through multi-hop message passing. However, existing methods often overlook the dynamic nature of evolving interactions, which impedes adaption to changing user preferences and distribution shifts newly arriving data. Thus, their scalability performances real-world environments are limited. In this study, we propose GraphPro, a framework that incorporates parameter-efficient graph pre-training with prompt...

Click-Through Rate (CTR) prediction, which aims to estimate the probability that a user will click an item, is essential component of online advertising. Existing methods mainly attempt mine interests from users' historical behaviours, contain directly interacted items. Although these have made great progress, they are often limited by recommender system's direct exposure and inactive interactions, thus fail all potential interests. To tackle problems, we propose Neighbor-Interaction based...

With the growth of high-dimensional sparse data in web-scale recommender systems, computational cost to learn high-order feature interaction CTR prediction task largely increases, which limits use models real industrial applications. Some recent knowledge distillation based methods transfer from complex teacher shallow student for accelerating online model inference. However, they suffer degradation accuracy process. It is challenging balance efficiency and effectiveness models. To address...