OneRanker: Unified Generation and Ranking with One Model in Industrial Advertising Recommendation

TL;DR

This work proposes OneRanker, achieving architectural-level deep integration of generation and ranking, and proposes input-output dual-side consistency guarantees, providing a new paradigm with industrial feasibility for generative advertising recommendations.

Abstract

The end-to-end generative paradigm is revolutionizing advertising recommendation systems, driving a shift from traditional cascaded architectures towards unified modeling. However, practical deployment faces three core challenges: the misalignment between interest objectives and business value, the target-agnostic limitation of generative processes, and the disconnection between generation and ranking stages. Existing solutions often fall into a dilemma where single-stage fusion induces optimization tension, while stage decoupling causes irreversible information loss. To address this, we propose OneRanker, achieving architectural-level deep integration of generation and ranking. First, we design a value-aware multi-task decoupling architecture. By leveraging task token sequences and causal mask, we separate interest coverage and value optimization spaces within shared representations, effectively alleviating target conflicts. Second, we construct a coarse-to-fine collaborative target awareness mechanism, utilizing Fake Item Tokens for implicit awareness during generation and a ranking decoder for explicit value alignment at the candidate level. Finally, we propose input-output dual-side consistency guarantees. Through Key/Value pass-through mechanisms and Distribution Consistency (DC) Constraint Loss, we achieve end-to-end collaborative optimization between generation and ranking. The full deployment on Tencent's WeiXin channels advertising system has shown a significant improvement in key business metrics (GMV - Normal +1.34\%), providing a new paradigm with industrial feasibility for generative advertising recommendations.

Figures (3)

• Figure 1: Comparison between existing methods and ours.
• Figure 2: The overall framework of OneRanker, consists of three stages: (i) the Generation stage which utilizes a generative backbone network pre-trained with multiple tokens based on user behavior sequences using heterogeneous tokens; (ii) the Multi-Task/Target-Aware stage which decouples interest and value objectives via task tokens and enables coarse-grained target awareness via fake item tokens; (iii) the Ranking stage which achieves fine-grained value alignment through a unified decoder with input-output consistency guarantees;
• Figure 3: Effectiveness of $\mathcal{L}_{\text{DC}}$: (a) Absolute rank difference of items between Step 2 and Step 3 at each rank for models with and without $\mathcal{L}_{\text{DC}}$; (b) Top-$K$ items overlap ratio between Step 2 and Step 3 with varying $K$ for the two models.