GateDiffInt: Gate-Mediated Controllable Diffusion and Multi-Intent LLM Distillation for User Behavior Modeling

2026-08-19

AI Summary

为解决用户行为序列中的噪声-意图耦合问题，提出GateDiffInt框架，通过可控扩散过程和双门控机制增强去噪，并利用大语言模型蒸馏多意图至学生模型。

Abstract

Existing ranking models encode intent only implicitly, making it hard to disentangle structured intents of varying strength and temporal scale. Noise and intent in behavior sequences are mutually reinforcing---we call this Noise--Intent Coupling (NIC). Noise dilutes true intents, while the lack of structured intent priors leaves denoising without a clear target.To address NIC, we propose GateDiffInt, an intent interaction framework for industrial ranking. It uses the final conversion signal to jointly align sequence denoising and intent extraction. GateDiffInt applies a controllable forward diffusion process with dual gating to enhance and denoise behavior sequences. A large language model then acts as teacher to distill four structured intents---long-term, short-term, latent, and conversion---into a lightweight student model. The enhanced sequence and structured intent representations are deeply fused via attention to produce intent-aware representations for conversion-rate prediction.Extensive experiments on public and large-scale industrial datasets show consistent gains over strong baselines. In online A/B tests serving hundreds of millions of daily active users, GateDiffInt delivers substantial GMV improvements and has been deployed to primary traffic, confirming both effectiveness and production readiness.

Problem

• Noise-Intent Coupling
• Ranking Models
• Behavior Sequences
• Structured Intents

Innovation

• Gate-Mediated Controllable Diffusion
• Multi-Intent LLM Distillation
• Noise--Intent Coupling (NIC)
• Structured Intents
• Attention Mechanism