This article analyzes Kuaishou’s May 2024 paper on LLM‑driven recommendation, detailing its dual‑tower architecture, contrastive learning of user and item embeddings, and a CVR‑auxiliary task that together improve cold‑start handling and boost both offline and online AUC metrics.
This article details NetEase Cloud Music's technical approach to building implicit user relationship chains—using SimHash, Item2Vec, and MetaPath2Vec embeddings, large‑scale vector search, and a unified service architecture—to address cold‑start challenges across multiple business scenarios.
This article presents a systematic study of recall techniques for new‑user cold‑start in content recommendation, describing a baseline two‑tower model, a Dual Attention Network (DAN) fusion approach, and an enhanced Contextual‑Gate DAN that dynamically balances content and product sequences, together with offline and online evaluation results and future directions.
