📝 Publications

SIGIR 2024
sym

Sequential Recommendation with Latent Relations based on Large Language Model
Shenghao Yang, Weizhi Ma, Peijie Sun, Qingyao Ai, et al.

  • Academic Impact: In this work, we leverage rich world knowledge which is compressed in LLM to discover latent item relations for enhanced recommendation.
WWW 2024 Industry
sym

Collaborative-Enhanced Prediction of Spending on Newly Downloaded Mobile Games under Consumption Uncertainty
Peijie Sun, Yifan Wang, Min Zhang, Chuhan Wu, et al.

  • Indudtry Impact: The research outcomes of this work achieved a 50.65% increase in online revenue on Huawei HiMedia, and an 18.43% uplift in traffic across Huawei-related business scenarios.
IEEE TKDE 2023
sym

Neighborhood-Enhanced Supervised Contrastive Learning for Collaborative Filtering
Peijie Sun, Le Wu, Kun Zhang, Xiangzhi Chen, Meng Wang.

PaperWithCode
Code

  • Academic Impact: Our proposed algorithm outperforms the best algorithm on the candidate matching leaderboard on BARS, significantly on Yelp and Gowalla datasets. Our proposed nearest neighbor-enhanced supervised contrastive module significantly outperforms the baseline model. On three real-world datasets, Yelp2018, Gowalla, and Amazon-Book, our model surpasses the original SGL by 10.09%, 7.09%, and 35.36% on NDCG@20, respectively.
CIKM 2023
sym

Understanding User Immersion in Online Short Video Interaction
Zhiyu He, Shaorun Zhang, Peijie Sun, Jiayu Li, Xiaohui Xie, Min Zhang, Yiqun Liu

  • Academic Impact: In this study, we primarily investigate users’ immersion states while browsing short videos. We argue that, instead of traditional metrics like user dwell time on short videos, immersion state should be considered as a more effective evaluation criterion for user satisfaction with short video content.
SIGIR 2019
sym

A Neural Influence Diffusion Model for Social Recommendation
Le Wu, Peijie Sun, Yanjie Fu, Richang Hong, Xiting Wang, Meng Wang

  • Academic Impact: We are among the first to apply graph convolutional techniques to social recommendation tasks. According to Google Scholar, as of today, our work has garnered over 500 citations. Besides, we have also provied the tensorflow and paddlepaddle version code.