Could you explain more of your dataset?

Hi,

   Thanks for your code and dataset. Is it convenient for you to explain more of the features as you mentioned:
Columns 3-102: The 100-dim flattened context; 10 features per arm (incorporating the content of the article and its match with the visiting user), first the features for arm 1, then arm 2, etc. up to arm 10.

   What are the meanings of 10 features per arm?

     