Edge 299: A Taxonomy to Understand Tool-Augmented Language Models
What are the different ways to augment LLMs with tools.
In this Issue:
Different types of tool-augmented LLM methods.
Google’s TALM research proposes a method to augment LLMs with differentiable tools.
Microsoft’s DeepSpeed-Chat to fine tune LLMs using RLHF.
💡 ML Concept of the Day: A Taxonomy to Understand Tool-Augmented Language Models
In the previous issue of this series, we introduced the concept of augmenting LLM capabilities with access to tools. There are different approaches to combine pretrained LLMs with information and actions from external tools and most of them still remain highly experimental. In general, we can organized tool-augmented LLM methods in two main groups: