Efficiently Deploying LLMs with Controlled Risk

Michael J. Zellinger, Matt Thomson

Deploying large language models in production requires simultaneous attention to efficiency and risk control. Prior work has shown the possibility to cut costs while maintaining si…