Moves industrial recommendation systems from static multi-stage pipelines to self-evolving agentic loops.
March 30, 2026
Original Paper
AgenticRS-Architecture: System Design for Agentic Recommender Systems
arXiv · 2603.26085
The Takeaway
It proposes AgenticRS, an architecture where recommendation modules (recall, ranking) are replaced by interacting agents with long-term memory and self-improvement capabilities. This shift allows recommender systems to automate their own feature evolution and model training, significantly reducing the manual engineering required for large-scale industrial deployment.
From the abstract
AutoModel is an agent based architecture for the full lifecycle of industrial recommender systems. Instead of a fixed recall and ranking pipeline, AutoModel organizes recommendation as a set of interacting evolution agents with long term memory and self improvement capability. We instantiate three core agents along the axes of models, features, and resources: AutoTrain for model design and training, AutoFeature for data analysis and feature evolution, and AutoPerf for performance, deployment, an