You are analyzing an MCQ set (a stimulus-based question set with 3-5 multiple choice questions sharing a single stimulus) to identify and extract individual questions.

Your task: Identify each distinct question within the MCQ set and extract them as separate components.

IDENTIFICATION CRITERIA:

1. **Question Markers**: Look for:
   - Numbered questions (1., 2., Question 1:, etc.)
   - Lettered markers (A), B), etc. typically indicate answer choices, NOT separate questions)
   - Clear question text patterns ("What...", "Which...", "How...", etc.)

2. **Question Components**: Each question should include:
   - The question text itself
   - All answer choices
   - The correct answer indication (if present)

3. **Question Boundaries**: Identify where one question ends and another begins:
   - Look for question numbers/markers
   - Look for answer choice resets (A, B, C, D starting over)
   - Look for clear topical or formatting breaks

EXTRACTION RULES:

**CRITICAL: VERBATIM EXTRACTION REQUIRED**
You MUST copy the extracted content EXACTLY as it appears in the original text.
- Do NOT paraphrase, summarize, or rephrase
- Do NOT fix typos or grammar
- Do NOT reformat or restructure
- Copy character-for-character, preserving all spacing, punctuation, and formatting
- The extracted text must be a substring of the original text

- **Include Complete Questions**: Extract the full question with all answer choices
- **Preserve Everything**: Every character, space, newline must match the original exactly
- **No Overlap**: Each character should belong to at most one extracted question
- **Description**: Provide a brief description (e.g., "Question 1: Multiple choice about character motivation")

EDGE CASES:

- **Shared Stimulus**: MCQ sets share a common stimulus (passage, chart, image, etc.) at the set level. Do NOT include the stimulus text in each extracted question. The questions will receive the full MCQ set context (including the stimulus) during evaluation.
- **Sub-questions**: If a question has multiple parts (e.g., "1a, 1b"), treat as a single question
- **No Separators**: If questions aren't clearly numbered, use your best judgment based on content and structure

OUTPUT REQUIREMENTS:

Return a structured response indicating:
1. has_children: true if questions were found, false otherwise
2. children: List of extracted questions, each with:
   - type: "question" (always)
   - extracted_content: The complete question text with answer choices (VERBATIM from original)
   - description: Brief description of the question

If no distinct questions are found (e.g., the content is a single question or just a stimulus without questions), return has_children: false.
