mlsanityData to 14 Oct 2026

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

arXiv 2603.01590 ↗First posted 2 Mar 2026cs.IR cs.LGcited by 2

Not on the map. Why?

Abstract (arXiv)

Content-driven platforms such as Xiaohongshu often leverage click-through rate (CTR) prediction models for recommendation. However, these models depend heavily on item ID embeddings, which perform poorly in item cold-start settings. In this paper, we present IDProxy, a production-scale system developed at Xiaohongshu to address this challenge. IDProxy leverages multimodal large language models (MLLMs) to generate proxy embeddings from rich 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. Extensive offline and online experiments demonstrate the effectiveness of the method, which has been deployed in 2025 in Xiaohongshu's Content Feed and Display Ads features, reaching hundreds of millions of users daily.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Builds on 31 of its 51 references that are in the map. First posted on arXiv 2 Mar 2026.

Builds on

31 of 51 references.

References · 31

  1. 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Taskscited by 2,565
  2. 2018Self-Attentive Sequential Recommendationcited by 1,184
  3. 2017Deep Interest Network for Click-Through Rate Predictioncited by 544
  4. 2017Opening the Black Box of Deep Neural Networks via Informationcited by 503
  5. 2019Linguistic Knowledge and Transferability of Contextual Representationscited by 383
  6. 2023A Survey on Large Language Models for Recommendationcited by 379
  7. 2025Layer by Layer: Uncovering Hidden Representations in Language Modelscited by 311
  8. 2015VBPR: Visual Bayesian Personalized Ranking from Implicit Feedbackcited by 151
  9. 2017Deep Learning based Recommender System: A Survey and New Perspectivescited by 136
  10. 2018Billion-scale Commodity Embedding for E-commerce Recommendation in Alibabacited by 105
  11. 2021Contrastive Learning for Cold-Start Recommendationcited by 79
  12. 2023Large Language Models for Generative Recommendation: A Survey and Visionary Discussionscited by 73
19 more
  1. 2019Warm Up Cold-start Advertisements: Improving CTR Predictions via Learning to Learn ID Embeddingscited by 66
  2. 2024NoteLLM-2: Multimodal Large Representation Models for Recommendationcited by 44
  3. 2021Deep Learning for Click-Through Rate Estimationcited by 33
  4. 2022Towards Understanding the Overfitting Phenomenon of Deep Click-Through Rate Prediction Modelscited by 33
  5. 2024Harnessing Multimodal Large Language Models for Multimodal Sequential Recommendationcited by 33
  6. 2024Enhancing Taobao Display Advertising with Multimodal Representations: Challenges, Approaches and Insightscited by 23
  7. 2021A Semi-Personalized System for User Cold Start Recommendation on Music Streaming Appscited by 21
  8. 2025Cold-Start Recommendation towards the Era of Large Language Models (LLMs): A Comprehensive Survey and Roadmapcited by 20
  9. 2024Large Language Models as Data Augmenters for Cold-Start Item Recommendationcited by 19
  10. 2023Contrastive Collaborative Filtering for Cold-Start Item Recommendationcited by 15
  11. 2022Alleviating Cold-start Problem in CTR Prediction with A Variational Embedding Learning Frameworkcited by 13
  12. 2024LLMTreeRec: Unleashing the Power of Large Language Models for Cold-Start Recommendationscited by 13
  13. 2021Cold Start Similar Artists Ranking with Gravity-Inspired Graph Autoencoderscited by 12
  14. 2024Large Language Model Simulator for Cold-Start Recommendationcited by 10
  15. 2025A Survey on Large Language Models in Multimodal Recommender Systemscited by 7
  16. 2025MOON: Generative MLLM-based Multimodal Representation Learning for E-commerce Product Understandingcited by 7
  17. 2023Test Time Embedding Normalization for Popularity Bias Mitigationcited by 4
  18. 2019Wasserstein Collaborative Filtering for Item Cold-start Recommendationcited by 3
  19. 2025A Metric for MLLM Alignment in Large-scale Recommendationcited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Music Playlist Captioning at Scale with Large Language Modelscited by 1
  2. 2026Privacy in Personalized AI Is a System Property, Not Just a Model Propertycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Addressing Cold-start Problem in Click-Through Rate Prediction via Supervised Diffusion Modeling
  2. 2024CELA: Cost-Efficient Language Model Alignment for CTR Prediction
  3. 2026Personalized LLM Recommendation with Semantic IDs: Autoregressive Beam Search vs. Multi-Token Prediction
  4. 2024Molar: Multimodal LLMs with Collaborative Filtering Alignment for Enhanced Sequential Recommendation
  5. 2025MMSRARec: Summarization and Retrieval Augumented Sequential Recommendation Based on Multimodal Large Language Model
  6. 2023UFIN: Universal Feature Interaction Network for Multi-Domain Click-Through Rate Prediction
  7. 2025DMGIN: How Multimodal LLMs Enhance Large Recommendation Models for Lifelong User Post-click Behaviors
  8. 2025Bridging Domain Gaps between Pretrained Multimodal Models and Recommendations