Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Generative Bid Shading in Real-Time Bidding Advertising

Yinqiu Huang, Hao Ma, Wenshuai Chen, Zongwei Wang, Shuli Wang, Yongqiang Zhang, Xue Wei, Yinhua Zhu, Haitao Wang, Xingxing Wang

Abstract

Bid shading plays a crucial role in Real-Time Bidding (RTB) by adaptively adjusting the bid to avoid advertisers overspending. Existing mainstream two-stage methods, which first model bid landscapes and then optimize surplus using operations research techniques, are constrained by unimodal assumptions that fail to adapt for non-convex surplus curves and are vulnerable to cascading errors in sequential workflows. Additionally, existing discretization models of continuous values ignore the dependence between discrete intervals, reducing the model's error correction ability, while sample selection bias in bidding scenarios presents further challenges for prediction. To address these issues, this paper introduces Generative Bid Shading (GBS), which comprises two primary components: 1) an end-to-end generative model that utilizes an autoregressive approach to generate shading ratios by stepwise residuals, capturing complex value dependencies without relying on predefined priors; and 2) a reward preference alignment system, which incorporates a channel-aware hierarchical dynamic network (CHNet) as the reward model to extract fine-grained features, along with modules for surplus optimization and exploration utility reward alignment, ultimately optimizing both short-term and long-term surplus using group relative policy optimization (GRPO). Extensive experiments on both offline and online A/B tests validate GBS's effectiveness. Moreover, GBS has been deployed on the Meituan DSP platform, serving billions of bid requests daily.

Generative Bid Shading in Real-Time Bidding Advertising

Abstract

Bid shading plays a crucial role in Real-Time Bidding (RTB) by adaptively adjusting the bid to avoid advertisers overspending. Existing mainstream two-stage methods, which first model bid landscapes and then optimize surplus using operations research techniques, are constrained by unimodal assumptions that fail to adapt for non-convex surplus curves and are vulnerable to cascading errors in sequential workflows. Additionally, existing discretization models of continuous values ignore the dependence between discrete intervals, reducing the model's error correction ability, while sample selection bias in bidding scenarios presents further challenges for prediction. To address these issues, this paper introduces Generative Bid Shading (GBS), which comprises two primary components: 1) an end-to-end generative model that utilizes an autoregressive approach to generate shading ratios by stepwise residuals, capturing complex value dependencies without relying on predefined priors; and 2) a reward preference alignment system, which incorporates a channel-aware hierarchical dynamic network (CHNet) as the reward model to extract fine-grained features, along with modules for surplus optimization and exploration utility reward alignment, ultimately optimizing both short-term and long-term surplus using group relative policy optimization (GRPO). Extensive experiments on both offline and online A/B tests validate GBS's effectiveness. Moreover, GBS has been deployed on the Meituan DSP platform, serving billions of bid requests daily.

Paper Structure

This paper contains 23 sections, 25 equations, 7 figures, 4 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Problem Definition
  5. Generative Model Preprocessing
  6. Method
  7. End-to-End Generative Model
  8. Reward Preference Alignment System
  9. Channel-Aware Hierarchical Dynamic Network (CHNet)
  10. Surplus Optimized Alignment
  11. Exploration Utility Alignment
  12. Experiments
  13. Experimental Setup
  14. Datasets
  15. Evaluation Metrics
  16. ...and 8 more sections

Figures (7)

  • Figure 1: Overview of bid shading process.
  • Figure 2: The winning rate and surplus curves derived from Meituan's non-ideal second-price auctions data.
  • Figure 3: The proposed generative model employs an encoder-decoder architecture to predict the token sequence autoregressively. It determines the specific value of each token using a decoding function $g(\cdot)$, then sums these values to obtain the final shading ratio. Additionally, it utilizes the gumbel softmax and teacher forcing strategies to ensure stability and accelerate convergence.
  • Figure 4: The architecture of CHNet. It outputs the PDF distribution through the hierarchical dynamic network.
  • Figure 5: The architecture of post-training with GRPO.
  • ...and 2 more figures