I digged into this problem this week. My conclusion is that there are 2 main performance issues for long thread page:
The initial loading time for opening the page is significantly longer than short threads.
Typing in followup input box is really laggy
The first one (longer initial loading time) still exists, but the second one has been fixed. Feel free to try!
Bohan
olive
•
Sep 14, 2024
Hi Nova,
This is a response adapted from our founder’s response to a similar post:
Deepest apologies to the frustration. Our team is still quite small, and we haven't yet implemented technologies like virtual scrolling to optimize page performance.
The situation is as follows: Currently, in LLM dialogue/search scenarios, there's an issue that can't be perfectly resolved, which is the model's memory problem. Due to the limited size of the LLM input window, it's unavoidable that if a conversation goes on for dozens or even hundreds of rounds, the model's contextual memory will be lost, leading to memory confusion.
Therefore, based on my personal experience, when finishing a discussion on a particular topic, starting a new thread for further questions tends to yield better and more accurate responses.
Given this experience, rather than optimizing the Long Thread page, we've been focusing more on how to guide users to start new threads for better answers. This is why we haven't done much performance optimization work on this page. I apologize for any inconvenience this may have caused.
Best, Olive
Nova
•
Sep 13, 2024
Like its super laggy
Anonymous User
•
Sep 13, 2024
+1
Bohan
Team•
Sep 19, 2024
Hi,
I digged into this problem this week. My conclusion is that there are 2 main performance issues for long thread page:
The initial loading time for opening the page is significantly longer than short threads.
Typing in followup input box is really laggy
The first one (longer initial loading time) still exists, but the second one has been fixed. Feel free to try!
Comments7
Bohan
Sep 19, 2024
Hi Nova,
I digged into this problem this week. My conclusion is that there are 2 main performance issues for long thread page:
The initial loading time for opening the page is significantly longer than short threads.
Typing in followup input box is really laggy
The first one (longer initial loading time) still exists, but the second one has been fixed. Feel free to try!
Bohan
olive
Sep 14, 2024
Hi Nova,
This is a response adapted from our founder’s response to a similar post:
Deepest apologies to the frustration. Our team is still quite small, and we haven't yet implemented technologies like virtual scrolling to optimize page performance.
The situation is as follows: Currently, in LLM dialogue/search scenarios, there's an issue that can't be perfectly resolved, which is the model's memory problem. Due to the limited size of the LLM input window, it's unavoidable that if a conversation goes on for dozens or even hundreds of rounds, the model's contextual memory will be lost, leading to memory confusion.
Therefore, based on my personal experience, when finishing a discussion on a particular topic, starting a new thread for further questions tends to yield better and more accurate responses.
Given this experience, rather than optimizing the Long Thread page, we've been focusing more on how to guide users to start new threads for better answers. This is why we haven't done much performance optimization work on this page. I apologize for any inconvenience this may have caused.
Best,
Olive
Nova
Sep 13, 2024
Like its super laggy
Anonymous User
Sep 13, 2024
+1
Bohan
Sep 19, 2024
Hi,
I digged into this problem this week. My conclusion is that there are 2 main performance issues for long thread page:
The initial loading time for opening the page is significantly longer than short threads.
Typing in followup input box is really laggy
The first one (longer initial loading time) still exists, but the second one has been fixed. Feel free to try!
Bohan