Researchers have developed ChronicleRec, a novel framework for pre-training and transferring user behavior models. This method compresses ultra-long historical action sequences into chronologically ordered "Chronicle Tokens," preserving recent behaviors while simplifying distant history. The framework uses a causal encoder and a multi-horizon design to learn complementary long-range interests and reconstruct recent behaviors from older history. Experiments on KuaiRand and Tencent AdLive datasets demonstrated ChronicleRec's superior performance compared to existing baselines, with a seven-day online A/B test confirming significant production gains. AI
IMPACT Enhances recommendation system efficiency by enabling effective modeling of long-term user behavior without prohibitive computational costs.
RANK_REASON This is a research paper detailing a new method for user modeling in recommendation systems. [lever_c_demoted from research: ic=1 ai=1.0]
Read on arXiv cs.IR (Information Retrieval) →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
