The main branch does not contain skipping of fragments that have been put in in the past, so the topic is always pulled back, even if an interesting topic grows out of it. - Diary 2023-08-26

Skip fragments put in the past = page memory. - page memory is difficult to see the value in the short term, but clearly beneficial in the long term - In the short term, in response to [Do you need 🤖MMR?” I thought - Diary 2023-08-26 - > The main branch does not contain skipping of fragments that have been put in in the past, so the topic is always pulled back, even if an interesting topic grows out of it. - Even if you put in MMR, you need this feature itself, or else the topic will be pinned. - image - For a human query 0, it is pulled to an existing sentence fragment and outputs 1. - If a human sees this and develops it into 2. - Without page memory, the same gravitational pull will occur again next time. - I thought the MMR was necessary to avoid one, but if there are multiple fragments of similarity, it’s natural to gravitate there once it’s well told. - It is not good for trade-offs between use and exploration if forces occur that keep you there. - So page memory is essential to care for this one. - Like this. - image

    - Extending arms in the direction of new clusters by not being pinned

This page is auto-translated from /nishio/話題のピン留め効果 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.