IntHQ: Task-Interactive Hierarchical Query on Dual-Stream Representations for Generative Recommendation

2026-08-10

AI Summary

This work addresses the triple collapse problem in multi-task generative recommendation—stemming from source signal dilution, rigid task relationship modeling, and insufficient exploitation of hierarchical features—by introducing a tri-component collaborative architecture. The proposed framework mitigates source collapse via dual-stream decoupling, overcomes relation collapse through adaptive task interaction modeling that captures dynamic dependencies, and alleviates hierarchy collapse using a hierarchical query mechanism that enables training-stage-aware aggregation of multi-granularity features. Supporting early task injection and input-adaptive fusion, the architecture consistently outperforms prevailing encoder-based approaches across four distinct task-head configurations. Upon deployment in Amap, the system achieved a 1.60% relative improvement in UVCTR.

Abstract

Multi-task learning over heterogeneous data is fundamental to modern recommendation, while generative models are emerging as the backbone of next-generation recommenders. However, the integration of multi-task learning into the generative paradigm remains largely unexplored. Existing multi-task recommenders, in both discriminative and generative paradigms, extract task-relevant features from a single task-agnostic representation and wire tasks into a predefined conversion funnel. We show that this scheme is inherently prone to a threefold collapse. Source collapse, where task-specific signals are injected late and diluted in the shared latent space. Relational collapse, where task dependencies are either implicitly absorbed by the backbone or statically fixed by predefined funnels. Hierarchical collapse, where tasks depend on features at different scales and shift across training stages. We propose IntHQ, a multi-task generative recommender with three components, each alleviating one collapse. Dual-Stream Decoupling (DSD) injects task identity into computation stream early and separates the shared context stream from the task-specific stream, alleviating signal dilution. Task-Interactive Modeling (TIM) replaces the predefined funnel with explicit cross-task interaction, letting each task condition on the realized outcomes of its predecessors with learned, input-adaptive strength. Hierarchical Querying (HQ) lets each task gather multi-scale information across different layers at different training stages. In offline evaluations, IntHQ consistently outperforms competitive encoder backbones under four representative task-head configurations. Deployed in production on Amap, serving hundreds of millions of users for travel recommendation, IntHQ yields a 1.60\% relative UVCTR lift.

Problem

• multi-task learning
• generative recommendation
• task collapse
• heterogeneous data
• task dependency

Innovation

• generative recommendation
• multi-task learning
• task-interactive modeling
• dual-stream representation
• hierarchical querying