Meituan deploys Generative Bid Shading to optimize ad spending

Researchers have developed Generative Bid Shading (GBS), a new approach for optimizing ad bidding in real-time advertising. GBS utilizes an autoregressive generative model to create shading ratios and a reward preference alignment system with a hierarchical dynamic network to extract features. This method aims to improve both short-term and long-term surplus by overcoming limitations of existing two-stage bid shading techniques. GBS has been successfully deployed on the Meituan DSP platform, handling billions of daily bid requests. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a novel generative approach for ad bidding that has been deployed at scale, potentially influencing industry practices.

RANK_REASON This is a research paper detailing a new method for ad bidding optimization.

Read on arXiv cs.LG →

• Generative Bid Shading
• Real-Time Bidding
• Meituan DSP
• CHNet
• GRPO
• arXiv

• paper
• product
• other