Short Video Recommendation Algorithm Frontier Research Forum at CCIR 2023

Recently, the 29th China Conference on Information Retrieval (CCIR 2023) was successfully held in Beijing, organized by the Chinese Society of Chinese Information Processing and co‑hosted by Tsinghua University, with Kuaishou sponsoring the event and co‑organizing a "Short Video Recommendation Algorithm Frontier Research" academic forum.

During the forum, Dr. Song Yang, Vice President of Kuaishou Technology, presented "Opportunities and Challenges of AI in Short Video" and introduced Kuaishou's cutting‑edge AI achievements in video content production, understanding, and distribution, as well as three open datasets: KuaiRec, KuaiRand, and KuaiSAR.

Dr. Song emphasized that short video is a crucial application scenario for recommendation technology and that releasing large‑scale recommendation datasets greatly benefits research on AI and recommendation algorithms.

The forum attracted more than 100 experts, scholars, and graduate students, featuring opening remarks by Dr. Jiang Peng, who highlighted the rapid growth of short‑video platforms, massive user interaction data, and research problems such as user retention, interest exploration, cold‑start, diversity, fairness, and large‑model recommendation.

Prof. Zhang Min from Tsinghua University delivered a keynote titled "Short Video Recommendation: Interdisciplinary Challenges and Opportunities," discussing how recommendation research intersects with economics, social science, psychology, and neuroscience, and presenting recent work on fair resource allocation and user immersion in short‑video browsing.

The forum included two technical sessions on reinforcement learning for user retention and user interest exploration, showcasing papers from Kuaishou and university partners that have appeared at top conferences such as SIGIR, KDD, NeurIPS, and CIKM.

All presentation slides are available for download for learning and exchange.

After the forum, more than 100 participants took a group photo together.

In the evening, Kuaishou's social‑science team and HR organized a campus recruitment dinner for over 70 students, where Dr. Song Yang welcomed the attendees and introduced the company's development, Q3 performance, and talent‑cultivation plans.

The Kuaishou Community Science team, the AI algorithm engine for core scenarios, provides end‑to‑end AI solutions for video, live streaming, e‑commerce, and local life, covering content understanding, recommendation, social interaction, and community growth, while collaborating with dozens of universities on top‑tier research that has produced award‑winning papers at SIGIR, ACM MM, CIKM, and other conferences.

Appendix:

[1] KuaiRec: A Fully‑observed Dataset for Recommender Systems – https://kuairec.com/

[2] KuaiRand: An Unbiased Sequential Recommendation Dataset with Randomly Exposed Videos – https://kuairand.com/

[3] KuaiSAR: A Unified Search and Recommendation Dataset – https://github.com/Ethan00Si/KuaiSAR

[4] Alleviating Matthew Effect of Offline Reinforcement Learning in Interactive Recommendation, SIGIR'23 – https://arxiv.org/abs/2307.04571

[5] KuaiSim: A Comprehensive Simulator for Recommender Systems, NeurIPS 2023 – https://neurips.cc/virtual/2023/poster/73528

[6] When Search Meets Recommendation: Learning Disentangled Search Representation for Recommendation – https://arxiv.org/abs/2305.10822

[7] TWIN: Two‑stage Interest Network for Lifelong User Behavior Modeling in CTR Prediction, KDD 2023 – https://arxiv.org/abs/2302.02352

[8] CCIR‑Kuaishou Academic Forum PPT download – https://pan.baidu.com/s/1bD1rA9U-5M9DEnPGbAzjtA?pwd=aq9i