Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GCRank: A Generative Contextual Comprehension Paradigm for Takeout Ranking Model

Ziheng Ni, Congcong Liu, Cai Shang, Yiming Sun, Junjie Li, Zhiwei Fang, Guangpeng Chen, Jian Li, Zehua Zhang, Changping Peng, Zhangang Lin, Ching Law, Jingping Shao

TL;DR

GCRank tackles the challenge of ranking in takeout advertising by modeling four contextual dimensions—static, dynamic, personalized, and collective—within a generative framework. It introduces the Generative Takeout Context Encoder (GCE) with PCE, DCE, and CCE, and the Generative Context Fusion (GCF) to produce unified, context-aware ranking signals. Extensive industrial-scale experiments and online A/B tests show significant improvements in CTR, CVR, and RPM, validating both the approach and its deployment in production. The work highlights a practical, scalable path for context-aware generative recommendations in complex LBS settings and provides insights into multi-granularity context modeling and efficient fusion through LoRA adapters.

Abstract

The ranking stage serves as the central optimization and allocation hub in advertising systems, governing economic value distribution through eCPM and orchestrating the user-centric blending of organic and advertising content. Prevailing ranking models often rely on fragmented modules and hand-crafted features, limiting their ability to interpret complex user intent. This challenge is further amplified in location-based services such as food delivery, where user decisions are shaped by dynamic spatial, temporal, and individual contexts. To address these limitations, we propose a novel generative framework that reframes ranking as a context comprehension task, modeling heterogeneous signals in a unified architecture. Our architecture consists of two core components: the Generative Contextual Encoder (GCE) and the Generative Contextual Fusion (GCF). The GCE comprises three specialized modules: a Personalized Context Enhancer (PCE) for user-specific modeling, a Collective Context Enhancer (CCE) for group-level patterns, and a Dynamic Context Enhancer (DCE) for real-time situational adaptation. The GCF module then seamlessly integrates these contextual representations through low-rank adaptation. Extensive experiments confirm that our method achieves significant gains in critical business metrics, including click-through rate and platform revenue. We have successfully deployed our method on a large-scale food delivery advertising platform, demonstrating its substantial practical impact. This work pioneers a new perspective on generative recommendation and highlights its practical potential in industrial advertising systems.

GCRank: A Generative Contextual Comprehension Paradigm for Takeout Ranking Model

TL;DR

GCRank tackles the challenge of ranking in takeout advertising by modeling four contextual dimensions—static, dynamic, personalized, and collective—within a generative framework. It introduces the Generative Takeout Context Encoder (GCE) with PCE, DCE, and CCE, and the Generative Context Fusion (GCF) to produce unified, context-aware ranking signals. Extensive industrial-scale experiments and online A/B tests show significant improvements in CTR, CVR, and RPM, validating both the approach and its deployment in production. The work highlights a practical, scalable path for context-aware generative recommendations in complex LBS settings and provides insights into multi-granularity context modeling and efficient fusion through LoRA adapters.

Abstract

The ranking stage serves as the central optimization and allocation hub in advertising systems, governing economic value distribution through eCPM and orchestrating the user-centric blending of organic and advertising content. Prevailing ranking models often rely on fragmented modules and hand-crafted features, limiting their ability to interpret complex user intent. This challenge is further amplified in location-based services such as food delivery, where user decisions are shaped by dynamic spatial, temporal, and individual contexts. To address these limitations, we propose a novel generative framework that reframes ranking as a context comprehension task, modeling heterogeneous signals in a unified architecture. Our architecture consists of two core components: the Generative Contextual Encoder (GCE) and the Generative Contextual Fusion (GCF). The GCE comprises three specialized modules: a Personalized Context Enhancer (PCE) for user-specific modeling, a Collective Context Enhancer (CCE) for group-level patterns, and a Dynamic Context Enhancer (DCE) for real-time situational adaptation. The GCF module then seamlessly integrates these contextual representations through low-rank adaptation. Extensive experiments confirm that our method achieves significant gains in critical business metrics, including click-through rate and platform revenue. We have successfully deployed our method on a large-scale food delivery advertising platform, demonstrating its substantial practical impact. This work pioneers a new perspective on generative recommendation and highlights its practical potential in industrial advertising systems.
Paper Structure (31 sections, 14 equations, 4 figures, 2 tables, 2 algorithms)

This paper contains 31 sections, 14 equations, 4 figures, 2 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Classical Ranking Models
  4. Generative Recommendation
  5. Preliminary
  6. Contextual Taxonomy
  7. Problem Formulation
  8. Methodology
  9. GCE: Dynamic Context Enhancer
  10. Knowledge Base Construction
  11. Online Retrieval and Representation Generation
  12. Optimized Inference Protocol
  13. GCE: Collective Context Enhancer
  14. Training Phase: AOI-Granularity Supervision
  15. Inference Phase: Multi-Granularity Generation
  16. ...and 16 more sections

Figures (4)

  • Figure 1: The overall framework of GCRank, consisting of two core modules: GCE (Generative Takeout Contextual Encoder) and GCF (Generative Contextual Fusion).
  • Figure 2: The detailed architecture of GCE and GCF.GCE is composed of three sub-modules: Personalized Context Enhancer (PCE), Dynamic Context Enhancer (DCE), and Collective Context Enhancer (CCE).
  • Figure 3: Hyperparameter studies of GCRank showing CTR AUC (left) and CVR AUC (right) performance across different configurations. Optimal configurations are highlighted in green.
  • Figure 4: The online deployment architecture of GCRank, featuring three specialized pipelines for real-time serving and generative context modeling.