Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

IDProxy: Cold-Start CTR Prediction for Ads and Recommendation at Xiaohongshu with Multimodal LLMs

Yubin Zhang, Haiming Xu, Guillaume Salha-Galvan, Ruiyan Han, Feiyang Xiao, Yanhua Huang, Li Lin, Yang Luo, Yao Hu

TL;DR

IDProxy is presented, a solution that leverages multimodal large language models (MLLMs) to generate proxy embeddings from rich content signals, enabling effective CTR prediction for new items without usage data.

Abstract

Click-through rate (CTR) models in advertising and recommendation systems rely heavily on item ID embeddings, which struggle in item cold-start settings. We present IDProxy, a solution that leverages multimodal large language models (MLLMs) to generate proxy embeddings from rich content signals, enabling effective CTR prediction for new items without usage data. These proxies are explicitly aligned with the existing ID embedding space and are optimized end-to-end under CTR objectives together with the ranking model, allowing seamless integration into existing large-scale ranking pipelines. Offline experiments and online A/B tests demonstrate the effectiveness of IDProxy, which has been successfully deployed in both Content Feed and Display Ads features of Xiaohongshu's Explore Feed, serving hundreds of millions of users daily.

IDProxy: Cold-Start CTR Prediction for Ads and Recommendation at Xiaohongshu with Multimodal LLMs

TL;DR

IDProxy is presented, a solution that leverages multimodal large language models (MLLMs) to generate proxy embeddings from rich content signals, enabling effective CTR prediction for new items without usage data.

Abstract

Click-through rate (CTR) models in advertising and recommendation systems rely heavily on item ID embeddings, which struggle in item cold-start settings. We present IDProxy, a solution that leverages multimodal large language models (MLLMs) to generate proxy embeddings from rich content signals, enabling effective CTR prediction for new items without usage data. These proxies are explicitly aligned with the existing ID embedding space and are optimized end-to-end under CTR objectives together with the ranking model, allowing seamless integration into existing large-scale ranking pipelines. Offline experiments and online A/B tests demonstrate the effectiveness of IDProxy, which has been successfully deployed in both Content Feed and Display Ads features of Xiaohongshu's Explore Feed, serving hundreds of millions of users daily.
Paper Structure (21 sections, 2 equations, 2 figures, 3 tables)

This paper contains 21 sections, 2 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Problem Formulation
  4. Related Work
  5. IDProxy: Proxy ID Embeddings with MLLMs
  6. Stage 1: MLLM-based Proxy Generation
  7. ID Embedding Preprocessing
  8. MLLM-Based Multimodal Encoding
  9. Proxy Alignment
  10. Stage 2: CTR-Aware Proxy Alignment
  11. Hierarchical Representation Partitioning
  12. Lightweight Multi-Granularity Adaptor for CTR
  13. End-to-End Joint Training with CTR Ranker
  14. Deployment
  15. Experimental Analysis
  16. ...and 6 more sections

Figures (2)

  • Figure 1: Visualizing item ID embeddings using t-SNE maaten2008visualizing. Left: MovieLens-1M embeddings learned by SASRec SASREC with feature crossing. Right: production embeddings from Xiaohongshu. Colors indicate item genres ("adventure", "sci-fi",...).
  • Figure 2: Overview of IDProxy's two-stage coarse-to-fine alignment framework, used on Xiaohongshu's Explore Feed (left).