Google

Strategy X: For each email, the system sends a long prompt that includes a detailed, 200-word definition of each category, several examples, and the full text of the customer email.
Strategy Y: For each email, the system sends a short prompt con

The inefficiency caused by long and complex natural language prompts is particularly acute in applications where a task must be performed repeatedly. In such scenarios, inputting the same lengthy prompt a large number of times leads to significant and unnecessary computational costs.

Inefficiency of Long Prompts in Repetitive Tasks

Analyze the following scenario and identify the primary source of computational inefficiency. Explain how the nature of the task exacerbates this problem.

LLM-Powered Email Categorization System

A company designs a customer service chatbot. For every user question, the system prepends a 500-word block of text containing detailed instructions and company policies. This combined text is then sent to a language model. Analyze the computational efficiency of this design for a system that handles thousands of user questions daily, and identify the primary source of the inefficiency.

Chatbot Prompting Strategy Analysis

A company is developing an automated system to classify thousands of customer support emails per day into one of three categories: 'Urgent', 'Standard', or 'Spam'. They are considering two prompting strategies for their large language model.

- **Strategy X:** For each email, the system sends a long prompt that includes a detailed, 200-word definition of each category, several examples, and the full text of the customer email.
- **Strategy Y:** For each email, the system sends a short prompt con

Learn Before

Related