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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Yubin, Xu, Haiming, Salha-Galvan, Guillaume, Han, Ruiyan, Xiao, Feiyang, Huang, Yanhua, Lin, Li, Luo, Yang, Hu, Yao
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910037772861440
author Zhang, Yubin
Xu, Haiming
Salha-Galvan, Guillaume
Han, Ruiyan
Xiao, Feiyang
Huang, Yanhua
Lin, Li
Luo, Yang
Hu, Yao
author_facet Zhang, Yubin
Xu, Haiming
Salha-Galvan, Guillaume
Han, Ruiyan
Xiao, Feiyang
Huang, Yanhua
Lin, Li
Luo, Yang
Hu, Yao
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2603_01590
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle IDProxy: Cold-Start CTR Prediction for Ads and Recommendation at Xiaohongshu with Multimodal LLMs
Zhang, Yubin
Xu, Haiming
Salha-Galvan, Guillaume
Han, Ruiyan
Xiao, Feiyang
Huang, Yanhua
Lin, Li
Luo, Yang
Hu, Yao
Information Retrieval
Machine Learning
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.
title IDProxy: Cold-Start CTR Prediction for Ads and Recommendation at Xiaohongshu with Multimodal LLMs
topic Information Retrieval
Machine Learning
url https://arxiv.org/abs/2603.01590