A team is using a language model to perform a classification task, but the results are not what they expected. Based on the provided case details, identify the primary flaw in the team's instruction to the model and explain why it is causing the poor performance.

Google

An example of a prompt for a classification task is: 'You are provided with a text. Please determine whether it discusses the risks of AI.' This instruction directs a Large Language Model to classify a document based on its content.

Example Prompt for an AI Risk Classification Task

A researcher needs to filter a large database of news articles to identify only those that discuss the potential dangers of artificial intelligence. Which of the following instructions would be the most effective and direct way to command a language model to perform this classification task?

Debugging a Classification Prompt

You are building a system to automatically sort customer feedback emails into two categories: 'Urgent Complaint' and 'General Inquiry'. Write a single, clear instruction (a prompt) that you would give to a language model to perform this classification task for any given email.

Learn Before

Related