Regarding token consumption

[Celina-love-sweet]

Hello, I would like to know if the token consumption mentioned in your paper (for example, 2.5K on the LOCOMO dataset, gpt-4o-mini) refers to the total consumption, including both memory generation and searching for memories based on the memory store to answer questions, or if it only refers to the consumption associated with searching for memories and answering questions based on the memory store.

[yuwang-xai]

2.5k should be the final memory length, rather than the total token consumption