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

Researchers have introduced RQ-GMM, a novel Residual Quantized Gaussian Mixture Model designed to improve multimodal semantic discretization in click-through rate (CTR) prediction systems, as announced in a paper published on February 20, 2026, on arXiv. This research addresses the persistent challenge of effectively incorporating multimodal content into CTR prediction models, which has been hampered by technical limitations in existing approaches. The paper identifies that directly incorporating continuous embeddings from pre-trained models into CTR models yields suboptimal results due to misaligned optimization objectives and convergence speed inconsistencies during joint training. By discretizing embeddings into semantic IDs before feeding them into CTR models, researchers have found a more effective solution, though previous methods have suffered from limited code efficiency and semantic representation capabilities. The RQ-GMM model represents a significant advancement in this area, offering improved performance in handling multimodal data for more accurate click-through rate predictions.