Billion-Scale Thumbnail Optimization for Uncurated Short-Form Videos via Multi-Armed Bandits

2026-10-04

AI Summary

This study addresses the display optimization challenge arising from the absence of manually curated thumbnails for massive video content on short-video platforms. We propose a billion-scale, real-time automated thumbnail selection framework that employs multi-stage candidate generation and deep visual quality models to filter the candidate set. Notably, this work pioneers the extension of online multi-armed bandit algorithms to O(B)-scale deployment, leveraging visual priors to effectively reduce exploration costs. Through a low-latency serving architecture, the system dynamically replaces default frames with optimal thumbnails. Global deployment results demonstrate statistically significant improvements in user discovery rates and core engagement metrics. Ultimately, this framework establishes a new paradigm for thumbnail optimization within large-scale industrial recommendation systems.

Abstract

This paper introduces a real-time thumbnail optimization system deployed at a global $O(B)$ scale on a major short-form video platform. Unlike traditional long-form content, where custom thumbnails are heavily curated by creators, a considerable fraction of short-form videos are published without human-selected artwork. To address this uncurated corpus, we present a fully automated, end-to-end framework that replaces static default frames with dynamic, data-driven selections across billions of videos. To the best of our knowledge, this is the first published work demonstrating an online Multi-Armed Bandit framework successfully deployed at an $O(B)$ scale for uncurated short-form video discovery. Our solution pairs a multi-stage candidate generation pipeline with a low-latency serving infrastructure. By initializing the exploration framework with image-specific priors derived from a deep visual quality model, the system minimizes exploration costs and dynamically serves optimal thumbnails at serving time. Global deployment demonstrates statistically significant improvements in core user discovery and engagement metrics.

Problem

• Thumbnail Optimization
• Short-Form Videos
• Uncurated Content
• Billion-Scale

Innovation

• Multi-Armed Bandits
• Thumbnail Optimization
• Short-Form Videos
• Deep Visual Quality Model
• Billion-Scale System