A Self-Triggered Agentic Push Recommendation System

2026-08-03

AI Summary

This work addresses the joint decision-making problem of “whether to push” and “when to push” in notification-based recommendation systems by proposing the first self-triggered end-to-end framework that enables dynamic, closed-loop global optimization under strict resource constraints. The approach employs a dual-agent architecture: a planning agent dynamically determines the next invocation time via gated ordinal regression, while an execution agent decides on push actions based on trajectory rewards. A lightweight filtering mechanism is further introduced to control computational overhead and ensure behavioral plausibility. Deployed in a live system, the method outperforms conventional fixed-frequency paradigms, yielding a 0.2843% increase in user active days, a 1.9089% reduction in push permission opt-out rate, and a 79.42% decrease in computational cost.

Abstract

Push notification is a critical recommendation scenario on large-scale platforms, allowing the system to proactively reach users outside the application to improve long-term re-engagement. However, designing an optimal push system requires handling a complex action space for the "whether and when" delivery problem under strict system resource constraints. Existing solutions typically fall into two passive paradigms: pre-planned frequency methods that allocate delivery times via offline modeling, limiting real-time adaptability; and fixed-interval triggering methods that periodically poll the system, creating a strict dilemma between excessive computational overhead and diminished optimal timing capture. Furthermore, such multi-stage frameworks severely suffer from local optima. To overcome these limitations, in this paper, we propose STEPS, a proactive, Self-Triggered End-to-end Agentic Push Recommendation System, which is already fully deployed at Douyin with over 1 billion users. STEPS reformulates push recommendation as a self-triggered agentic process in which the system decides not only whether to send a push, but also when to invoke itself again, thereby forming a closed loop that balances real-time effectiveness and efficiency. Specifically, STEPS consists of two decision transformer-based agents: a planning agent that schedules the next system invocation using a gated ordinal regression method, and an execution agent that decides whether to send a push based on trajectory rewards. Furthermore, we introduce a lightweight filtering agent to both control computational overhead and act as a crucial safeguard against unreasonable planning behaviors. Online A/B testing demonstrates that STEPS significantly increases user active days by 0.2843% and reduces the push permission disablement rate by 1.9089%, while the filtering agent reduces computational overhead by 79.42%.

Problem

• push recommendation
• self-triggered system
• action space
• resource constraints
• re-engagement

Innovation

• Self-Triggered
• Agentic Recommendation
• Decision Transformer
• Push Notification System
• End-to-End Optimization