Taming Large Language Models* - An Input-Output Perspective (*and a side trip to FeatureLanguage)

Abstract

This talk consists of two parts. The first part presents the automatic generation of complete applications suitable for undergraduate course MDE projects from a minimal, high-level DSL called FeatureLanguage. The second part introduces two frameworks for LLMs to improve the performance of LLM-based applications: the first integrates domain-specific workflows into LLM-based applications, whereas the second integrates domain-specific constraints into LLM-based applications.