Abstract:Display advertising provides significant value to advertisers, publishers, and users. Traditional display advertising systems utilize a multi-stage architecture consisting of retrieval, coarse ranking, and final ranking. However, conventional retrieval methods rely on ID-based learning to rank mechanisms and fail to adequately utilize the content information of ads, which hampers their ability to provide diverse recommendation lists. To address this limitation, we propose leveraging the extensive world knowledge of LLMs. However, three key challenges arise when attempting to maximize the effectiveness of LLMs: "How to capture user interests", "How to bridge the knowledge gap between LLMs and advertising system", and "How to efficiently deploy LLMs". To overcome these challenges, we introduce a novel LLM-based framework called LLM Empowered Display ADvertisement REcommender system (LEADRE). LEADRE consists of three core modules: (1) The Intent-Aware Prompt Engineering introduces multi-faceted knowledge and designs intent-aware <Prompt, Response> pairs that fine-tune LLMs to generate ads tailored to users' personal interests. (2) The Advertising-Specific Knowledge Alignment incorporates auxiliary fine-tuning tasks and Direct Preference Optimization (DPO) to align LLMs with ad semantic and business value. (3) The Efficient System Deployment deploys LEADRE in an online environment by integrating both latency-tolerant and latency-sensitive service. Extensive offline experiments demonstrate the effectiveness of LEADRE and validate the contributions of individual modules. Online A/B test shows that LEADRE leads to a 1.57% and 1.17% GMV lift for serviced users on WeChat Channels and Moments separately. LEADRE has been deployed on both platforms, serving tens of billions of requests each day.

A Simple Yet Strong Approach for Installation Prediction in ShareChat Ads

Click-Through Rate Prediction Algorithm Based on Modeling of Implicit High-Order Feature Importance

A Simple and Robust Ensemble for Click-Through Rate Prediction.

RecSys Challenge 2023: From data preparation to prediction, a simple, efficient, robust and scalable solution

Weighted Multi-Level Feature Factorization for App ads CTR and installation prediction

Context-aware advertisement recommendation for high-speed social news feeding

A Clickthrough Rate Prediction Algorithm Based on Users’ Behaviors

Exploiting Structural and Temporal Influence for Dynamic Social-Aware Recommendation

SSLRec: A Self-Supervised Learning Framework for Recommendation

Secure Social Recommendation Based on Secret Sharing

Deep Situation-Aware Interaction Network for Click-Through Rate Prediction.

Short Video-based Advertisements Evaluation System: Self-Organizing Learning Approach

Lightweight Boosting Models for User Response Prediction Using Adversarial Validation

Rec4Ad: A Free Lunch to Mitigate Sample Selection Bias for Ads CTR Prediction in Taobao

Composite Social Network for Predicting Mobile Apps Installation

Ads Recommendation in a Collapsed and Entangled World

LEADRE: Multi-Faceted Knowledge Enhanced LLM Empowered Display Advertisement Recommender System

Social4Rec: Distilling User Preference from Social Graph for Video Recommendation in Tencent

Moment&Cross: Next-Generation Real-Time Cross-Domain CTR Prediction for Live-Streaming Recommendation at Kuaishou

A Deep Behavior Path Matching Network for Click-Through Rate Prediction

Personalized Attraction Enhanced Sponsored Search with Multi-task Learning