Meta's Generative Ads Model (GEM)

Abstract

We’re sharing details about Meta’s Generative Ads Recommendation Model (GEM), a new foundation model that delivers increased ad performance and advertiser ROI by enhancing other ads recommendation models’ ability to serve relevant ads. GEM’s novel architecture allows it to scale with an increasing number of parameters while consistently generating more precise predictions efficiently. GEM propagates its learnings, leveraging a suite of post-training techniques across the entire ads model fleet, enabling a paradigm shift in Meta’s Ads Recommendation system. GEM leverages enhanced training scalability that efficiently utilizes thousands of GPUs for building and iterating an LLM-scale ads foundation model. GEM is already driving significant increases in ad conversions across Instagram and Facebook.