Google

A method for generating initial prompts is to instruct a Large Language Model (LLM) by providing it with a human-written description of the task. While this approach is direct and straightforward, it is dependent on the availability of a human-provided task description.

LLM-based Prompt Generation from a Task Description

An example template to directly instruct a Large Language Model to produce a prompt based on a task description is: `You are given a task to complete using LLMs. Please write a prompt to guide the LLMs. *task-description*`. This method provides a direct description of the task to the model.

Example of an Instruction for LLM-based Prompt Generation

Analyze the following scenario. What is the most probable cause of the poor-quality prompts, and what specific details should the team add to their task description to generate more effective and targeted initial prompts?

Troubleshooting Initial Prompt Generation

A startup wants to quickly develop an AI feature for classifying customer support tickets. They have a detailed classification guide written by their support manager, but no AI specialists on the team. They decide to generate their initial set of prompts by feeding the classification guide into a large language model. Analyze the primary strength and the main potential weakness of this specific approach for their situation.

Evaluating a Prompt Generation Strategy

A team is tasked with creating a system to summarize legal documents. To begin, they write a detailed, two-page description of the desired summary format, key information to include, and the target audience. They then input this description into a large language model with the instruction: 'Based on the provided document, generate a set of effective prompts for a summarization task.' When critiquing this team's strategy for creating an initial set of prompts, what is the most critical factor tha

Learn Before

Related