I suppose that you need org-mode to use Babel features (code and/or code execution results intermixed with your paper's text).
I have had trouble with .dot exporting using the builtin .dot exporter, but can't, for the life of me, remembering if my troubles were cross- reference or citations-related (an academic paper usually needs both...). I noted that the references done via John Kitchin's excellent org-ref didn't export to .dot. As far as I can tell, ox-pandoc does its own numbering (generating an intermediate temporary file that I never managed to catch) and does not use pandoc-crossref. ISTR that a "clever" use of org-ref allows to number equations... in LaTeX output. I also STR that it doesn't work in .docx output. John Kitchin (the org-ref author) thinks that the key is to generate a correct \LaTeX file and to export that to .docx via pandoc. At this point, writing a \LaTeX file with pandoc export in mind is probably your best bet. There exist (limited) alternatives to the Babel features of org-mode: - knitr (supports R and Python code in "session"mode, i. e. continuity between code chunks), - SageTeX (post-processing in a single Sagemath run ; can include code for a variety of math software, including R), - Pythontex (postprocessor, supports Python, Sage, Octave and a couple other, but not R ; a clever use of knitr and Pythontex is possible). Possible alternative: Markdown + Codebraid (supports a hanful of langiages, including R and (development branch) Sage). This should accept pandoc-crossref numbering. But it's Markdown, with limited formatting abilities... So, we have an array of partial solutions, none universal. Pick your poison... Out of curiosity: in which domain still exist journal publishers not accepting \LaTeX ? I suspect medicine... HTH, -- Emmanuel Charpentier
