Mishandling of whitespaces and formatting in uploaded documents

The whitespaces and formatting are not handled correctly when they are present in files and documents. I attach one file containing a tree structured defined in a loose and casual text based format. Every model has difficulty parsing and understanding it because the formatting gets removed.

Now if I base64 encode the above data, and then tell the models to first decode and then try to parse, they are able to parse it correctly.

I tested the plain text format with commercial services like ChatGPT and Grok. They are able to parse the tree correctly, just like the models in VeniceAI when given with base64 encoded version.

This mishandling of spaces occur with prompts and system prompts too most likely.

The prompt I use with plaintext version “fmt-test.txt” above is :

The file fmt-test.txt contains a tree where each node is a number. The tree is described using a loose and casual text based format. Try to parse the tree and show to me what was parsed. If you fail to parse then state the reason for failure.