Causal-Informed Hybrid Online Adaptive Optimization for Ad Load Personalization in Large-Scale Social Networks

This paper presents CTRCBO (Cohort-Based Trust Region Contextual Bayesian Optimization), a hybrid framework designed for personalizing ad load in large-scale social networks like Meta.

Key Contributions:

• Hybrid Framework: Combines Primal-Dual methods with Bayesian Optimization (BO) to maintain stability while allowing for efficient exploration.
• Causal Integration: Utilizes upstream Causal ML models to inform Gaussian Process Regression (GPR) surrogates.
• Scalability: Validated on a billion-user network, demonstrating faster convergence and improved personalization metrics.

Presented at the NeurIPS 2025 Workshop on Constrained Optimization for Machine Learning (COML).

[Full Paper on arXiv]