Analyze the sequence of tool calls generated by the model in the following scenario. Identify the primary inefficiency in the model's first step and explain why the third and fourth steps represent a more direct and correct use of the available tools for the user's full request.

Google

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

When a Large Language Model generates a specific string formatted as a tool call, it triggers an execution process. The model runs the designated tool, such as a web search, with the provided query. The output from this tool then replaces the original tool call string in the model's context. This newly inserted information is used by the LLM in subsequent prediction steps to formulate a correct and contextually aware answer.

Executing Tool Calls and Integrating Results in LLMs

A user asks a large language model, 'What is the current stock price of ExampleCorp (ticker: EXM)?'. The model has access to a tool designed to fetch stock prices which requires a company's ticker symbol. Which of the following represents the most appropriate and correctly structured command the model would generate to use this tool?

Evaluating LLM Tool Use Efficiency

A user prompts a large language model with the question: 'What is my next meeting today?'. The model has access to a tool named `calendar_lookup` which can search a user's calendar. This tool requires a single parameter called `search_term` to function. Based on the user's question, write the specific, structured command the model should generate to use this tool.

Learn Before

Related