← Back to the feed
AI & softwareResearch findings

TAGR: Temporally Adaptive Generative Recommendation for Live-Streaming Ads

Original published 2026.08.25 12:47 KSTAdded to Blackdeck 2026.10.20 03:09 KST

What has changed?

Temporal adaptation is introduced at three levels: live-ad tokenization, intent modeling, and preference alignment

Key points

  1. A temporally adaptive generative recommendation framework is proposed for live-stream advertising with rapidly changing content
  2. Adaptive tokenization, intent modeling, and preference alignment are implemented to track evolving live ads and user behavior
  3. Live-room entry and shopping-cart click rates increased by 8.5% and 7.4% respectively on a large-scale platform

Connect it to your work

The model dynamically adjusts to rapidly changing live content and promoted products

What remains unproven

This research note is based on an abstract; check the original for test conditions and application scope.

Originals used for this note

Links open the original publisher in a new tab.

  1. arxivTAGR: Temporally Adaptive Generative Recommendation for Industrial Live-Streaming Advertising