MMQ-v2: Align, Denoise, and Amplify: Adaptive Behavior Mining for Semantic IDs Learning in Recommendation

2025-10-29

AI Summary

In industrial recommendation systems, traditional item IDs suffer from poor scalability, while purely content-driven semantic item IDs (SIDs) fail to model dynamic user behaviors, leading to severe noise interference and imbalanced signal weighting in long-tail scenarios. To address these issues, we propose ADA-SID: (1) an adaptive behavior-content alignment mechanism that suppresses collaborative noise in long-tail items while preserving multimodal semantic integrity; (2) a dynamic behavior router that differentially weights SIDs based on behavioral information richness, thereby enhancing the expressiveness of critical behavioral signals; and (3) a hybrid quantization architecture (MMQ-v2) for joint modeling of multimodal item content and heterogeneous behavioral data. Extensive experiments on both public and large-scale industrial datasets demonstrate that ADA-SID consistently outperforms state-of-the-art methods on both generative and discriminative recommendation tasks, achieving superior generalization and accuracy.

Technology Category

• Cognitive Modeling & Cognitive Systems: Adaptive Behavior
• Data Mining & Knowledge Management: Recommender Systems
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Industrial recommender systems rely on unique Item Identifiers (ItemIDs). However, this method struggles with scalability and generalization in large, dynamic datasets that have sparse long-tail data.Content-based Semantic IDs (SIDs) address this by sharing knowledge through content quantization. However, by ignoring dynamic behavioral properties, purely content-based SIDs have limited expressive power. Existing methods attempt to incorporate behavioral information but overlook a critical distinction: unlike relatively uniform content features, user-item interactions are highly skewed and diverse, creating a vast information gap in quality and quantity between popular and long-tail items. This oversight leads to two critical limitations: (1) Noise Corruption: Indiscriminate behavior-content alignment allows collaborative noise from long-tail items to corrupt their content representations, leading to the loss of critical multimodal information. (2)Signal Obscurity: The equal-weighting scheme for SIDs fails to reflect the varying importance of different behavioral signals, making it difficult for downstream tasks to distinguish important SIDs from uninformative ones. To tackle these issues, we propose a mixture-of-quantization framework, MMQ-v2, to adaptively Align, Denoise, and Amplify multimodal information from content and behavior modalities for semantic IDs learning. The semantic IDs generated by this framework named ADA-SID. It introduces two innovations: an adaptive behavior-content alignment that is aware of information richness to shield representations from noise, and a dynamic behavioral router to amplify critical signals by applying different weights to SIDs. Extensive experiments on public and large-scale industrial datasets demonstrate ADA-SID's significant superiority in both generative and discriminative recommendation tasks.

Problem

• Address noise corruption in content-behavior alignment
• Resolve signal obscurity through dynamic weighting of SIDs
• Improve semantic ID learning for sparse long-tail items

Innovation

• Adaptive behavior-content alignment prevents noise corruption
• Dynamic behavioral router amplifies critical signals
• Mixture-of-quantization framework integrates multimodal information