Tools and Function Calling
As we discussed in Chapter 1, large language models (LLMs) have demonstrated outstanding reasoning abilities. Nevertheless, there’s not enough evidence that scaling models alone will be sufficient to “achieve high performance on challenging tasks such as arithmetic, commonsense, and symbolic reasoning” [1].
In this chapter, we will discuss some modifications in the ways we interact with LLMs that demonstrate a large improvement in such complex tasks.
We’ll start with discussing advanced prompt techniques, such as chain-of-thought (CoT), before covering what tools and functions are in the context of LLMs. We’ll investigate examples of giving a model information about recent events and the outside world by giving it access to Google Search and improving its mathematical capabilities with a calculator. Finally, we’ll discuss different ways of defining your custom tools on LangChain and briefly discuss the ReACT...
