Welcome to DimRed!

Ready to fine-tune your prompts? Here's how to get started.

Step 1: Data
Provide your data as a JSON or CSV file. Each entry should have an 'input' and an 'expected' key.
[
  {
    "input": {
      "text": "John Doe was seen fleeing the scene in a blue Toyota."
    },
    "expected": {
      "is_perpetrator": True
    }
  },
  {
    "input": {
      "text": "Jake Smith was the victim of a robbery that happened in NYC last fall."
    },
    "expected": {
      "is_perpetrator": False
    }
  }
]
Download Example Data
Step 2: Prompt
Write a prompt that will be tested against your data, or let us generate a starting point for you.
# Role
You are an expert at classifying perpetrators for a crime.

# Task
You will be given a text and you will need to classify whether the person in the text is a perpetrator or not.

# Instructions
Step 1. Carefully review the 'text' field in the input data.
Step 2. ...
Download Example Prompt
Step 3: Metric
Choose a metric to evaluate performance. This will be used to optimize your prompt during tuning.

Tuning Performance

Download Example Metric