Analyze the following interaction between a user and a language model that has access to a real-time financial data tool. Identify the primary reason the model failed to provide the live stock price and explain your reasoning.

Google

An example of how to prompt a Large Language Model for tool use involves providing both an instruction and a specific question. For instance, to find out where the 2028 Olympics will be held using a web search tool, the LLM can be given the following complete prompt:

`Your task is to answer the following question. You may use external tools, such as web search, to assist you.`

`Where will the 2028 Olympics be held?`

Example of a Prompt for LLM Tool Use

Following a prompt that enables tool use, a Large Language Model can generate a specific, structured command to execute an external tool. For instance, to answer the question about the 2028 Olympics, the model might produce the following tool call: `{tool: web-search, query: "2028 Olympics"}`.

Example of an LLM-Generated Tool Call

A user wants to ask a language model, which has access to a live weather tool, for the current temperature in London. Which of the following prompts is best structured to ensure the model uses its tool to get the real-time information?

A developer is building an application where a language model needs to access a company's internal employee database to answer questions. The developer uses the following prompt to find an employee's email address: `Find the email address for John Smith.` The developer observes that the model often fails to use the database tool and instead makes up an answer or states it cannot access private information. Analyze the provided prompt. What key component is missing that would make it more effective at triggering the use of the external database tool? Rewrite the prompt to improve its reliability.

Learn Before

Related