A structured prompt that helps you design a canary release plan for a new model version by asking the model to show its reasoning, not just its conclusion.
AI Evaluation and Testing
Prompts for planning evaluations, regression tests, and adversarial testing of AI systems.
Turns a short description into a concrete plan to design a plan for evaluating fairness across user demographics, with assumptions called out instead of silently guessed.
A ready-to-use prompt to design a plan for testing a model's behavior on out-of-scope requests, producing a structured draft you can act on directly rather than a vague suggestion.
Helps you design a plan for evaluating a search ranking model quickly, while still surfacing the open questions a real version of this work would need answered.
A structured prompt that helps you design a plan for testing a chatbot's escalation-to-human logic by asking the model to show its reasoning, not just its conclusion.
Turns a short description into a concrete plan to design a plan for evaluating long-context retention, with assumptions called out instead of silently guessed.
A ready-to-use prompt to design a plan for testing a model's numerical reasoning, producing a structured draft you can act on directly rather than a vague suggestion.
Helps you design a plan for evaluating multi-modal output quality quickly, while still surfacing the open questions a real version of this work would need answered.
A structured prompt that helps you design a plan for testing a system prompt against jailbreak attempts by asking the model to show its reasoning, not just its conclusion.
Turns a short description into a concrete plan to design a plan for evaluating a model's instruction-following, with assumptions called out instead of silently guessed.
A ready-to-use prompt to design a plan for testing data leakage in model outputs, producing a structured draft you can act on directly rather than a vague suggestion.
Helps you design a plan for evaluating a voice assistant's intent recognition quickly, while still surfacing the open questions a real version of this work would need answered.