Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The paper introduces IDProxy, a method to address the cold-start problem in CTR prediction by using multimodal large language models (MLLMs) to generate proxy embeddings for new items based on their content. These proxy embeddings are explicitly aligned with existing ID embeddings and optimized end-to-end with the ranking model, allowing for seamless integration into production systems. Experimental results and online A/B tests on Xiaohongshu demonstrate the effectiveness of IDProxy in improving CTR prediction for cold-start items in both Content Feed and Display Ads.
Solve the cold-start problem in CTR prediction by using multimodal LLMs to generate proxy embeddings that align with existing ID embeddings, enabling effective ranking of new items without usage data.
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.