I've not tried editing the summaries, but have edited the much larger /mem output as well as adding lore to the lore text files, mostly just to see if I can stop characters from using the same words over and over again or falling in love with some meaningless detail that only they care about.

I honestly don't know if it helps. The way "memories" work with textual AI makes no sense when compared to actual memory. Even if the lore and memory is spot on, the AI won't understand how to contextualize a passage (or when to consult it in the first place) with any coherence. Despite a concept being repeated throughout the lore and mem files, I still have to remind the AI characters about things.

I've also used the Reminder feature with mixed results-- mostly not worth using. At times the AI will start reciting the reminder. If you put ([AI]: Reminder: Don't say "But tell me," or "But hey," anymore.) in its reminder field, it might immediately say something like:

"AI Guy makes sure not to say 'But tell me' or 'But hey,' knowing how much this irritates the MC. 'But inform >me, what do you like to do for fun?' AI Guy asks.

... or something annoying like that.

I've also never actually seen my lore file being used even after chats reach 2,000 lines, but maybe I'm missing something.

I'm not saying these things do nothing, I just haven't noticed what these features actually do. I think it's just the limitations of the current model. If you are able to do any of the above and see results, please share.