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

2026-03-03

AI Summary

This work addresses the misalignment, procedural fragmentation, and information loss between ad generation and ranking in advertising recommendation systems by proposing OneRanker, the first framework to enable end-to-end co-optimization of both stages. OneRanker unifies the two tasks through a value-aware multi-task decoupling architecture, a coarse-to-fine target-aware mechanism, and dual consistency constraints on both input and output. Key technical innovations include task-specific token sequences, causal masking, fake item tokens, a ranking decoder, key/value passthrough, and a distributional consistency loss. The method has been fully deployed in Tencent’s WeChat advertising system, achieving a 1.34% increase in GMV, thereby demonstrating its industrial efficacy and deployability.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Search and Optimization: Algorithm Configuration
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

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.

Problem

• generative recommendation
• advertising recommendation
• generation-ranking misalignment
• target-awareness
• business value alignment

Innovation

• unified generation and ranking
• value-aware multi-task decoupling
• target awareness mechanism
• distribution consistency constraint
• end-to-end collaborative optimization