Empromptu Docs
  • Welcome
  • Getting Started
    • Take a Tour of Empromptu
    • How do you define Accuracy?
    • Quickstart
      • Quickstart - Archive
  • Advanced Configuration
  • No App, Use Prompt Optimizer
  • Examples
  • Basics
    • Projects and Tasks
    • Prompts Overview
      • Manual Optimization
      • Automatic Optimization
    • Input Optimization
    • Edge Case Optimizer
    • Evaluations
    • Models
    • Data Generator
    • Core Concepts
    • Usage Guide
      • Data Viewer
      • Performance Over Time
      • Prompt Explorer
      • Show Input Space
      • Prompt Performance
      • Prompt Playground
    • Best Practices
    • FAQ
    • Troubleshooting
Powered by GitBook
On this page
  1. Getting Started

How do you define Accuracy?

PreviousTake a Tour of EmpromptuNextQuickstart

Last updated 2 months ago

You can define accuracy how ever you want! In our experience, "accuracy is very use case specific". So what is accurate to one application, may not be accurate to another.

If you are building an AI chat bot, 92% accuracy may be good but if you are building a AI cybersecurity tool, that acts on behalf of humans, you may need to work towards 99%.

You can customize your definitions through our evaluations feature in 4 ways:

  1. Empromptu's Standard Templates

  2. User Created (that's you)

  3. End user confirmed - If you have a thumbs up or thumbs down or a subject matter expert doing human in the loop

  4. Generated - If you see an example in Empromptu's platform, you can tag that example and add it as an evaluation.

Want to get started creating an evaluation?

Visit our Evaluations to learn more