ACT: Automated Constraint Targeting for Multi-Objective Recommender Systems

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chang, Daryl, Wu, Yi, She, Jennifer, Wei, Li, Heldt, Lukasz
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916931830808576
author Chang, Daryl
Wu, Yi
She, Jennifer
Wei, Li
Heldt, Lukasz
author_facet Chang, Daryl
Wu, Yi
She, Jennifer
Wei, Li
Heldt, Lukasz
contents Recommender systems often must maximize a primary objective while ensuring secondary ones satisfy minimum thresholds, or "guardrails." This is critical for maintaining a consistent user experience and platform ecosystem, but enforcing these guardrails despite orthogonal system changes is challenging and often requires manual hyperparameter tuning. We introduce the Automated Constraint Targeting (ACT) framework, which automatically finds the minimal set of hyperparameter changes needed to satisfy these guardrails. ACT uses an offline pairwise evaluation on unbiased data to find solutions and continuously retrains to adapt to system and user behavior changes. We empirically demonstrate its efficacy and describe its deployment in a large-scale production environment.
format Preprint
id arxiv_https___arxiv_org_abs_2509_03661
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ACT: Automated Constraint Targeting for Multi-Objective Recommender Systems
Chang, Daryl
Wu, Yi
She, Jennifer
Wei, Li
Heldt, Lukasz
Information Retrieval
Machine Learning
Recommender systems often must maximize a primary objective while ensuring secondary ones satisfy minimum thresholds, or "guardrails." This is critical for maintaining a consistent user experience and platform ecosystem, but enforcing these guardrails despite orthogonal system changes is challenging and often requires manual hyperparameter tuning. We introduce the Automated Constraint Targeting (ACT) framework, which automatically finds the minimal set of hyperparameter changes needed to satisfy these guardrails. ACT uses an offline pairwise evaluation on unbiased data to find solutions and continuously retrains to adapt to system and user behavior changes. We empirically demonstrate its efficacy and describe its deployment in a large-scale production environment.
title ACT: Automated Constraint Targeting for Multi-Objective Recommender Systems
topic Information Retrieval
Machine Learning
url https://arxiv.org/abs/2509.03661