A tool for checking when AI outputs keep changing
can give different results even when the same instruction is used again. PromptProbe is built to make that problem easier to inspect. It runs the same prompt several times and shows where the answers differ.
The goal is to help people building with LLMs spot unstable output before it becomes a product or problem. The maker is looking for direct from people who use LLMs in their own work.
Key points
- The tool runs the same prompt more than once.
- It compares the results and marks where they differ.
- It targets people building with LLMs.
- The main problem is unstable output from the same instruction.