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

2026-03-02

AI Summary

This work addresses the cold-start problem in advertising and recommendation systems, where new items suffer from inaccurate click-through rate (CTR) prediction due to a lack of historical user interactions. To tackle this challenge, the authors propose a novel approach based on multimodal large language models (MLLMs) that leverages content signals to generate proxy embeddings. These embeddings are explicitly aligned with the traditional ID embedding space for the first time and are jointly optimized end-to-end with the ranking model under the CTR objective. The proposed method is seamlessly integrated into a large-scale online ranking system and has been deployed in both the Explore Feed content recommendation and display advertising modules of Xiaohongshu, serving hundreds of millions of users daily. Both offline experiments and online A/B tests demonstrate significant performance improvements.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web 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.

Problem

• cold-start
• CTR prediction
• item ID embeddings
• recommendation systems
• advertising

Innovation

• cold-start
• CTR prediction
• multimodal LLMs
• proxy embeddings
• embedding alignment