Prompt Suggestions
Learn how to take advantage of automatically generated suggestions to improve your prompts based on evaluation results.
What are Prompt Suggestions?
Prompt Suggestions are automatically generated recommendations to improve your prompt based on your latest evaluation results. They help you identify areas where your prompt can be enhanced to achieve better outcomes.
Whenever you run a prompt, from the playground, the API or a dataset, that has an evaluation connected, you’ll see suggestions appear next to the Refiner button in the prompt editor.
Latitude analyzes the latest evaluation results to identify patterns and potential improvements. Based on this analysis, we generate specific suggestions that can help you:
- Improve response quality
- Address common failure points
- Enhance clarity and effectiveness
- Optimize for specific metrics
- Fix identified issues
Getting Started
To start using prompt suggestions:
- Create evaluations for your prompt
- Run evaluations in batch mode or enable live evaluation
- Collect enough evaluation results for meaningful analysis
- Check the suggestions tab for improvement recommendations
- Implement the most relevant suggestions
- Run evaluations again to measure improvement
Remember that suggestions are most effective when you have a good amount of evaluation data to work with. We recommend running evaluations on at least 20-30 responses before checking for suggestions.