Evidence · first-party tested · observation #53
Grouped column headers in scanned tables lose relationships during extraction; column semantics drift.
An image-only scanned research paper created from an OCR-bearing original by converting pages to images and rebuilding the PDF without searchable text. It contains text, multi-column layouts, figures, charts, tables, captions, and references, and is designed to stress OCR and reconstruction quality.
Extracted treatment table with diameter measurements.
Scanned bar chart of tree mortality by year and treatment.
The observation
Grouped column headers in scanned tables lose relationships during extraction; column semantics drift.
Criterion: nested header handling · Scenario: Scanned multi-column research paper (group: scanned-research-paper)
Query this
get_evidence({
tool: "llamaparse",
scenario: "scanned-research-paper"
})MCP · admin.futuresmart.ai/api/mcp
Free with attribution.
Free with attribution.
Same input, other tools
LlamaParse ✓ WorkedLlamaParse ✓ WorkedLlamaParse ✓ WorkedLlamaParse ✓ WorkedLlamaParse ✓ WorkedLlamaParse ✓ WorkedLlamaParse ✓ WorkedAudio Studio ✗ FailedAudio Studio ◐ MixedAudio Studio ✓ WorkedAudio Studio ✓ WorkedAudio Studio ✓ WorkedAudio Studio ✓ WorkedAudio Studio ✓ WorkedAudio Studio ✓ Worked
Real outputs, no retouching · every cell queryable via API & MCP · aidemos.com