DaQuaMRec
Keynotes & Talks
Embedded Files
Guillaume Salha-Galvan
Shanghai Jiao Tong University
Multimodal LLMs in Action: Addressing Cold-Start Recommendation at Xiaohongshu, a Leading Chinese Social Media Platform
INVITED TALK
Abstract
Xiaohongshu, also known internationally as RedNote, is a leading Chinese content-driven social and e-commerce platform with more than 300 million users. The platform hosts a massive and continuously evolving stream of posts, images, and short videos spanning a wide range of topics, making large-scale recommender systems central to its operation.
Like many content-driven platforms, Xiaohongshu relies on click-through rate (CTR) prediction models to rank and recommend content. These models, however, depend heavily on item ID embeddings and typically struggle when newly uploaded items have little or no interaction history. This talk presents IDProxy, a production-scale system designed at Xiaohongshu to address this cold-start challenge.
IDProxy leverages multimodal large language models (LLMs) to generate proxy embeddings from rich multimodal content signals, enabling CTR prediction for new items in the absence of usage data. Through a lightweight coarse-to-fine mechanism, these proxies are aligned with the ID embedding space and trained end-to-end with the ranking model, allowing seamless integration into production-facing pipelines.
The talk discusses the motivations behind IDProxy, its design and evaluation at scale, and the industry lessons learned from bringing multimodal LLMs into a large-scale recommender system. IDProxy is now deployed across Xiaohongshu's Content Feed and Display Ads features, where it supports recommendations reaching hundreds of millions of users every day.
This talk is based on work selected for an oral presentation in the Industry Track at RecSys 2026.
Bio
Guillaume Salha-Galvan is an Associate Professor at Shanghai Jiao Tong University (SJTU). He is a faculty member of the SJTU Paris Elite Institute of Technology, an international institute established at SJTU in partnership with four leading French engineering schools: École Polytechnique, Mines Paris, Télécom Paris, and ENSTA. He is also the founder and principal investigator of the Sino-French Data and AI Lab.
His research focuses on recommender systems, large language models, and their real-world applications. He is a regular author and Program Committee member at RecSys, where three of his past papers were shortlisted for awards in 2020, 2021, and 2024. Several of his research contributions have also had significant real-world impact, notably by powering recommender systems used by millions of users.
His research is supported by several competitive grants and talent programs, including the NSFC Excellent Young Scientists Fund (Overseas), the Siyuan Young Scholar Award, the Shanghai Magnolia Talent Program, and the Xiaomi Young Scholar Award, for which he was the first non-Chinese recipient.
Prior to joining SJTU, he gained nearly a decade of industry experience, including serving as Director of Research at Kibo Ryoku and as a Research Scientist and Coordinator for Music Recommendation at Deezer. He received his Ph.D. in Computer Science from École Polytechnique in France.
Page updated
Google Sites
Report abuse