Eight-Folio Reproducible Sampler of the Voynich Manuscript

[tavie]

AI is used here mainly as a tool for spotting patterns and checking consistency, it doesn’t generate the transcriptions or translations on its own

LLMs currently are not identifying patterns in the Voynich.  They are hallucinating patterns when you prompt them, to keep you engaged. We have a You are not allowed to view links. Register or Login to view. of the forum for these kinds of outputs.  Take a look at the post showing why we do this, and also at some of the threads we've moved there.

These other threads also involve LLMs claiming to have identified hidden patterns or themes.  While the classifications of the patterns vary, their outputs tend to have common features
-no proof provided to establish the existence of patterns, only a statement of their existence 
-assertion that the results are reproducible, yet the methodology is unfollowable 
-it's often structured under headings expected for an academic paper but there's just no meat in there or any kind of argument propelling the thesis to the conclusion
-as a result, often we have no idea what it's going on about, and it's clear the person presenting the findings also doesn't understand them
-occasionally, the LLM makes up Voynichese and then "translates" its own nonsense

Speaking of the later, "otolshade"?  I've looked for a handful of the words your LLM has listed for 33v, and I've not found any of them. The list of words is also identical between 33v and 85r, but the "translation" is completely different.

Every EVA assignment in the sampler follows clear, reproducible rules, so anyone can trace it back to the original folio.

Nothing in the paper allows for reproducibility.  We don't know how the LLM assigns a W number, nor why it gives different translations for the same words in different folios, not to mention where it is finding some of these words...