A few methods I use to construct the ATTG for the memory field.

ATTG is short for [ Author: ; Title: ; Tags: ; Genre: ]. If you are not familiar with ATTG, I recommend you first read the memory guide by pume_ (pume_ calls ATTG 'Tags') and then come back here.

What The Memory Block looks like

If you read Pume_'s guide, this should be clear, but just to avoid confusion, this is what a full memory block looks like:

[ Author: <Name of author(s)>; Title: <Story Title - Titlecase>; Tags: <tag(s) - lowercase>; Genre: <genre(s) - lowercase> ]
<Metaplot: Summary of story (preferably in "backcover blurb style")>
***

Stuff in <> are placeholders, an actual memory block would not have any <> You can leave out any element of the memory block, it is fine if the ATTG only contains [ Tags: godzilla ] and nothing else, for example.

Example Of Memory Blocks:

Full block:

...

Title Case Or lowercase?

Should This Go Into Tags: Or Genre:?

Occasionally, it can be unclear, if a keyword belongs in Tags: or in Genre:. In that case, token probabilities can usually help.

Say you are unsure where to put superheroes. This is how to figure it out:

0. (Optional: For the prompt, provide a story outline to boost the relevant keyword. For example, Batman and Superman are fighting against Mario and Luigi!.)
1. Pump up the bias for superheroes. (so that it shows up in the probability viewer)
2. Add [ Tags: to the prompt. Note down the before token probabilities for the beginning of the keyword. (superhero in this example)
3. Replace [ Tags: with [ Genre:. Note down the before token probabilities for the beginning of the keyword.
4. Compare token probabilities. Higher token probability wins. (In this example, superhero shows higher probability under Tags: than under Genre: so we would use it for Tags:)