Please Clarification on OCR layout labels used in OCR_LAYOUT_PROMPT

Hi,

I am using the OCR_LAYOUT_PROMPT to generate HTML output with layout-aware OCR annotations. The model returns several predefined layout labels, but I could not find clear documentation describing their exact meaning and intended usage.

I am planning to use this output as training data for fine-tuning an OCR/layout understanding model, so I need to ensure that the labels are correctly interpreted and consistently applied.

Could you please clarify what each of the following labels represents in practice and how they should be used in downstream processing?

Caption Footnote Equation-Block List-Group Page-Header Page-Footer Image Section-Header Table Text Complex-Block Code-Block Form Table-Of-Contents Figure Chemical-Block Diagram Bibliography Blank-Page

Specifically, I would appreciate clarification on:

The semantic differences between similar categories (e.g., Image vs Figure, or Text vs Complex-Block) Whether these labels are mutually exclusive or hierarchical Whether overlapping regions are expected and how they should be handled Whether this label set follows an existing dataset taxonomy (e.g., PubLayNet, DocLayNet) or is custom-defined

This clarification is particularly important for me because I am preparing a dataset for fine-tuning an OCR/layout understanding model, and I want to ensure label consistency and correctness during training data generation.

Any documentation, guidelines, or recommendations would be greatly appreciated.

Thanks!