Complex tables with page breaks produce hallucitating results

Bug

The extracted table contains misplaced column names and corrupt data parsing pipeline.

Steps to reproduce

Source Image screenshot: (data obfuscated to not contain PHI)

Docling version

2.65.0

Python version

3.14

Pipeline Options used:

pipeline_options.do_table_structure = True pipeline_options.table_structure_options.do_cell_matching = True pipeline_options.table_structure_options.mode = TableFormerMode("accurate") pipeline_options.do_ocr = False // document has embedded text

Example of the markdown output using vlm:

file.md

Example of the markdown output using dl_parse_v4:

file2.md

Screenshot from rendered markdown:

All data is syntetic. No personal information shared