Yapay Zeka · 4 dk okuma · 27 Nisan 2026
KuaiLive: First Real-Time Live Streaming Recommendation Dataset
Researchers release a 21-day interaction log from Kuaishou covering 23,772 users and 452,621 streamers to enable dynamic recommendation research.
Kaynak: arxiv/cs.AI · Changle Qu, Sunhao Dai, Ke Guo, Xiao Zhang, Liqin Zhao, Shijun Wang, Yannan Niu, Lantao Hu, Han Li, Jun Xu · orijinali aç ↗ ↗
KuaiLive is the first public dataset capturing real-time live streaming interactions, enabling research into dynamic recommendation systems.
- — Dataset spans 21 days from Kuaishou, covering 23,772 users and 452,621 streamers with precise timestamps.
- — Records multiple interaction types: clicks, comments, likes, gifts—reflecting actual user engagement patterns.
- — Includes rich metadata for users and streamers, enabling realistic simulation of evolving candidate pools.
- — Supports diverse tasks: top-K recommendation, click prediction, watch time, gift pricing, and fairness research.
- — Addresses gap in academic research caused by absence of public live streaming benchmarks.
- — Enables multi-behavior and multi-task learning approaches previously difficult to validate.
- — Published openly to accelerate industry-academic collaboration on streaming recommendation.
Sık sorulanlar
- KuaiLive is the first publicly available dataset capturing real-time interactions on a live streaming platform (Kuaishou). It records 23,772 users and 452,621 streamers over 21 days, including clicks, comments, likes, and gifts. It matters because live streaming recommendation differs fundamentally from traditional systems—content is dynamic, user intent shifts rapidly, and engagement is immediate. Prior academic research lacked public benchmarks reflecting these dynamics, so KuaiLive enables researchers to develop and validate algorithms on realistic data.