> ## Documentation Index
> Fetch the complete documentation index at: https://distillkitplus.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Formatters

## Available Formatters (`dataset.format_function`)

<div className="overflow-x-auto">
  <table className="min-w-full divide-y divide-gray-200 dark:divide-gray-700">
    <thead>
      <tr>
        <th className="px-4 py-3 text-left text-sm font-medium text-gray-500 dark:text-gray-400 uppercase tracking-wider">Formatter Name</th>
        <th className="px-4 py-3 text-left text-sm font-medium text-gray-500 dark:text-gray-400 uppercase tracking-wider">Description</th>
        <th className="px-4 py-3 text-left text-sm font-medium text-gray-500 dark:text-gray-400 uppercase tracking-wider">Input Format</th>
        <th className="px-4 py-3 text-left text-sm font-medium text-gray-500 dark:text-gray-400 uppercase tracking-wider">Output Format</th>
      </tr>
    </thead>

    <tbody className="divide-y divide-gray-200 dark:divide-gray-700">
      <tr>
        <td className="px-4 py-3 text-sm text-gray-900 dark:text-gray-100">`default_format`</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">Standard chat format</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">Examples with `messages` containing chat turns (role and content)</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">Formatted chat using the tokenizer's template</td>
      </tr>

      <tr>
        <td className="px-4 py-3 text-sm text-gray-900 dark:text-gray-100">`sharegpt_format`</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">For ShareGPT-style data</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">Data with `conversations` containing turns (from: human/gpt/system, value)</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">Standard chat format with system message if missing</td>
      </tr>

      <tr>
        <td className="px-4 py-3 text-sm text-gray-900 dark:text-gray-100">`comparison_format`</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">For comparing two responses</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">Data with `prompt`, `response_a`, `response_b`, `rationale`, and `winner`</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">Structured chat showing comparison and result</td>
      </tr>

      <tr>
        <td className="px-4 py-3 text-sm text-gray-900 dark:text-gray-100">`format_for_tokenization`</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">Simple text extraction</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">Any data</td>
        <td className="px-4 py-3 text-sm text-gray-600 dark:text-gray-300">Just the text content or full example if no text field</td>
      </tr>
    </tbody>
  </table>
</div>

Notes:

* All formatters except `format_for_tokenization` use the tokenizer's chat template
* `default_format` is used if an unknown formatter is specified