OpenAI Publishes GPT Prompt Engineering Guide

prompt-engineeringopenaigpt-4llmfew-shot-learning

Abstraction: OpenAI's six-strategy guide for eliciting better GPT-4 responses

Key points:

Six high-level strategies: write clear instructions, provide reference text, split complex tasks into subtasks, give the model time to think, use external tools, and test changes systematically
Tactics include using the system message for persona/instructions, chain-of-thought prompting, and recursive summarization — drawn from published LLM research
"Use external tools" tactic suggests generating Python code for math rather than doing calculations directly in the model; code must be run in a sandbox
Systematic testing is aided by the OpenAI Evals framework for comparing prompt variants against gold-standard answers
Other providers (Microsoft Azure, Google Gemini) have released similar guides; community debate exists over long-term value of prompt engineering as models improve

Connections: Openai · GPT-4 · Prompt Engineering · Large Language Models · Chain Of Thought