GenRecs Tokenization: Transforming Ads Recommendation at Meta with Stable Model Vocabulary

Meta has introduced GenRecs, a generative framework that shifts its advertising systems from traditional score-based matching to a model that uses large language models to reason about ad content and user preferences. By utilizing a stable model vocabulary, the system aims to improve ad targeting precision by better understanding the context of both the user and the advertisement. This update marks a transition toward integrating generative AI reasoning directly into Meta's recommendation infrastructure to enhance outcome predictions for advertisers.