Prompt Chaining

A technique for interacting with large language models where multiple prompts are issued in sequence, with each prompt building on the output of the previous one to achieve a more refined or accurate result. In accessibility and bias mitigation contexts, prompt chaining enables self-reflection workflows where a model first generates a response, then is asked to examine that response for biases or stereotypes, and finally rewrites its answer. This approach leverages the model's ability to critique and improve its own outputs without requiring model retraining. Prompt chaining is used in applications ranging from bias mitigation and content moderation to accessible content generation and multi-step reasoning tasks.