RQ-GMM: Residual Quantized Gaussian Mixture Model for Multimodal Semantic Discretization in CTR Prediction

Researchers have introduced RQ-GMM, a new AI model that significantly improves click-through rate (CTR) prediction for online ads and recommendations. The model uses a novel 'Residual Quantized Gaussian Mixture Model' to better process multimodal content like images and text by converting them into discrete semantic IDs. In online A/B tests on a major short-video platform, it achieved a 1.502% gain in Advertiser Value and is now fully deployed, serving daily recommendations for hundreds of millions of users.