How to Train Your Large Language Model

Abstract

Contents include: There is a demand to use large language models to aid, enhance, and automate current workflows; Large Language Models are a compilation of trainings; Project Mayflower is engineering a Large Language Model (LLM); Our government customers want to know how they can use LLMs to benefit low and high side endeavors; Our customers' needs conflict with qualities of most LLMs; Mayflower's ultimate goal is solving high side capability; Using a proxy problem, we have a FR research question that will guide our high side research question; We have a prototype!; We know our LLM is learning something; We also know our LLM can be creative; We also developed a tool for side-by-side comparison; LLMs have numerous AI Engineering challenges; Semantic search is search + prompt engineering; We want to find the "best" way; "Best" should include RAI; Next steps include high side transition and more; In conclusion, Mayflower is in full bloom.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jul 19, 2023
Accession Number
AD1207046

Entities

People

  • Shannon Gallagher

Organizations

  • Carnegie Mellon University

Tags

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Big Data
  • Copyrights
  • Department Of Defense
  • Edge Computing
  • Engineering
  • Engineers
  • Governments
  • Guarantees
  • High Altitude
  • Human Rights
  • Infrastructure
  • Intelligence Analysts
  • Language
  • Materials
  • Models
  • Reinforcement Learning
  • Software Development
  • Training
  • Universities

Fields of Study

  • Computer science

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Oceanography.
  • Software Engineering.