AI4Editing/MagicQuill · Sizing and prompts matter?

When describing, or prompting for the desired result, I have found that sizing details are giving better results, so when clicking the green tick, does that help the AI Platform to record and improve the resulting images? Does it learn from these positive responses? Equally, when you place a red X because the result was not desired, does it also learn from such responses? From what I have seen so far, it seems to learn and we get less inaccurate representations! Maybe it is learning more about what we desire in the resulting image? Quality prompting is also helping I suspect!