CarperAI is a democratized AI research team that enhances the performance of preference learning. It offers a CARP model that produces embeddings of passages and reviews with high similarity, as well as a CARP-CoOp model that uses context optimization to tune reviews. The team also offers diff models trained on millions of commits scraped from GitHub. According to the website, these models help developers better understand code changes and data scientists gain insights into customer preferences. CarperAI differentiates itself from competitors by offering a Story-Critique dataset that provides an information-rich measure of preferences on story content.