Bidding for Influence: Auction-Driven Diffusion Image Generation

Lillian Sun, Henry Huang, Fucheng Warren Zhu, Giannis Daras, Constantinos Costis Daskalakis

Keywords: Diffusion Models, Online Advertising, Auction Theory

Abstract: We introduce the first generative image auction, utilizing diffusion models to express multi-agent preferences. Our mechanism first composes advertisers’ bids and text prompts inside the diffusion model’s reverse process through a dynamic score composition. Then, it implements Monte Carlo sampling to calculate VCG-based payments and select high-welfare images. Extensive experiments on a two-agent benchmark demonstrate three core properties: (1) bid monotonicity, (2) efficiency improvement of up to 21% higher welfare than a single-winner VCG baseline, and (3) approximate incentive compatibility, with average regret achieving below 10% when deviating from truthful bidding. Crucially, these benefits are achieved while preserving high image quality. Our study establishes a principled bridge between auction theory and controllable image diffusion, laying a foundation for economically-aligned, multi-stakeholder image generation in advertising and beyond.

Submission Number: 26