Extract verbatim spans from the documents for each placeholder in the template.

Question: {{ question }}

Template to fill:
{{ template }}

Placeholders to extract for:
{{ placeholder_spec }}

Documents:
{{ docs_text }}

Instructions:
1. For each placeholder, find EXACT verbatim quotes from the documents
2. Copy text exactly - no paraphrasing or modification
3. For each span, include which document it came from (0-indexed)
4. Return a JSON object mapping placeholder names to arrays of objects with "text" and "doc" fields
5. If no relevant information for a placeholder, use an empty array

Return ONLY valid JSON like:
{
  "METHODOLOGY": [{"text": "exact quote about methods...", "doc": 0}],
  "RESULTS": [{"text": "exact quote about results...", "doc": 1}]
}