Tractable Hierarchical Control of Autoregressive Language Models

Max Scribner, Antonio Vergari, Vaishak Belle

Constraining the generation of autoregressive large language models (LLMs) is an important component of integrating language models into formal systems. In the generation of code a…