Benchmarking document-conversion quality and fidelity

[StephanMeijer]

Does anyone know of tools, frameworks, or established methods for benchmarking document-conversion quality?

I’m particularly interested in evaluating how faithfully a conversion such as DOCX → Markdown preserves document structure, including headings and section hierarchy, tables, ordered and unordered lists, links, emphasis, inline formatting, textual content, and logical reading order, etc.

Ideally, I’m looking for a way to compare the converted output against a reference and produce a meaningful score or structured diff, rather than relying entirely on manual review.

Are there existing benchmarks, metrics, test suites, or recommended approaches for this?