A new research paper introduces Memento, a personalized retrieval-augmented framework designed to improve recommendation systems by effectively scaling long-term user history data. Memento addresses challenges like attention dilution and catastrophic forgetting by treating historical user engagements as a document corpus and ad requests as queries, using Maximal Marginal Relevance (MMR) for retrieval. The framework includes two applications: Representation Memento for feature augmentation and Data Memento for multipass training. Memento demonstrates significant resource efficiency gains (5-10x) and low latency (sub-10ms), leading to a 1% CTR lift and 1.2% CVR lift in production for Facebook Feed and Reels, enabling personalization with over a year of user history. AI
IMPACT Enhances personalization in recommendation systems by effectively utilizing extensive historical user data, potentially improving user engagement and conversion rates.
RANK_REASON Publication of an academic paper detailing a new framework and its performance metrics. [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 →
