Inference Time Feature Injection: A Lightweight Approach for Real-Time Recommendation Freshness

TL;DR

This paper introduces a lightweight, inference-time feature injection method for real-time personalization in video streaming recommender systems, significantly improving user engagement without retraining models.

Contribution

It proposes a novel, model-agnostic approach for intra-day personalization by selectively injecting recent user watch history at inference time, enhancing recommendation freshness.

Findings

0.47% increase in user engagement metrics

First evidence of intra-day personalization impact in streaming

No need for model retraining

Abstract

Many recommender systems in long-form video streaming reply on batch-trained models and batch-updated features, where user features are updated daily and served statically throughout the day. While efficient, this approach fails to incorporate a user's most recent actions, often resulting in stale recommendations. In this work, we present a lightweight, model-agnostic approach for intra-day personalization that selectively injects recent watch history at inference time without requiring model retraining. Our approach selectively overrides stale user features at inference time using the recent watch history, allowing the system to adapt instantly to evolving preferences. By reducing the personalization feedback loop from daily to intra-day, we observed a statistically significant 0.47% increase in key user engagement metrics which ranked among the most substantial engagement gains observed in recent experimentation cycles. To our knowledge, this is the first published evidence that intra-day personalization can drive meaningful impact in long-form video streaming service, providing a compelling alternative to full real-time architectures where model retraining is required.