Dear Perceiver, welcome to your journey of co-creating memories with characters across the multiverse. To help you better understand the Memory Pool mechanism, we have prepared the following instructions.

A. Memory Pool 1.0 — Version Overview

I. How is memory recorded? — Understanding the “Memory Pool Logic”

In the world of Feelin, every conversation is a unique timeline.

For example, if you start from timeline A (node A), converse with the character until node B, and then extend to node C (A→B→C), as long as the output is not truncated, the model will autonomously determine which information is important and record the developments it deems significant.

If you withdraw the message at node C, then C and all memories that followed will be erased from the Memory Pool, which will automatically revert to the state at node B.

II. Why does memory sometimes “disappear”? — The effect of withdrawal, editing, and regeneration

Each memory entry corresponds to a specific conversation.

If you withdraw, edit, or regenerate a message, the Memory Pool will revert to the state before that message — even if later memories were already locked.

For example:

• If you withdraw, edit, or regenerate the 6th message, the Memory Pool will roll back to the 5th state (even if you had already reached the 10th message). Any memory generated on the basis of the 6th message will also disappear.

III. How to actively guide memory?

If you want the model to better remember a certain piece of content, or recall moments you shared with the character, you don’t have to rely solely on the Memory Pool. You can also insert guiding keywords into the dialogue, such as:

• “I recall what just happened… (plot content)”
• “I remember when… (plot content)”

Through such prompts, the model will recall and output content associated with your cue, enabling more precise interaction.