Beyond Quality: Unlocking Diversity in Ad Headline Generation with Large Language Models

2025-08-26

AI Summary

Advertising headline generation suffers from an imbalance between quality and diversity: existing approaches typically optimize either linguistic quality or click-through rate (CTR) in isolation, leading to homogeneous outputs. This paper proposes DIVER, a framework that jointly optimizes semantic diversity and generation quality within a single forward pass. Its core contributions are: (1) a dual-aware data generation pipeline that models both semantic and stylistic dimensions; and (2) a multi-stage, multi-objective optimization strategy integrating supervised fine-tuning with reinforcement learning. Extensive experiments on real-world industrial data demonstrate that DIVER achieves significant improvements—+4.0% in advertising value (ADVV) and +1.4% in CTR—while simultaneously maintaining high output quality and diversity. To our knowledge, DIVER is the first method to enable efficient, joint optimization of both objectives in advertising headline generation.

Technology Category

• Search and Optimization: Learning to Search
• Natural Language Processing: Generation
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

The generation of ad headlines plays a vital role in modern advertising, where both quality and diversity are essential to engage a broad range of audience segments. Current approaches primarily optimize language models for headline quality or click-through rates (CTR), often overlooking the need for diversity and resulting in homogeneous outputs. To address this limitation, we propose DIVER, a novel framework based on large language models (LLMs) that are jointly optimized for both diversity and quality. We first design a semantic- and stylistic-aware data generation pipeline that automatically produces high-quality training pairs with ad content and multiple diverse headlines. To achieve the goal of generating high-quality and diversified ad headlines within a single forward pass, we propose a multi-stage multi-objective optimization framework with supervised fine-tuning (SFT) and reinforcement learning (RL). Experiments on real-world industrial datasets demonstrate that DIVER effectively balances quality and diversity. Deployed on a large-scale content-sharing platform serving hundreds of millions of users, our framework improves advertiser value (ADVV) and CTR by 4.0% and 1.4%.

Problem

• Balancing quality and diversity in ad headline generation
• Overcoming homogeneous outputs from current CTR-optimized models
• Generating diverse high-quality headlines in single forward pass

Innovation

• Multi-stage multi-objective optimization framework
• Semantic- and stylistic-aware data generation pipeline
• Joint optimization for diversity and quality