Quick Answer performance regression

Dustin

Just some feedback that the updates to Quick Answers seems to have negatively affected performance. The text is slow to appear and it's also quite wordy in a way that is antithetical to how I feel Quick Answer should be.

Maybe the model needs to be tuned to be a bit more concise and performance can be improved so the answers really are quick 🌩️

Thibaultmol

Could you maybe provide some other examples? in this case you're just asking the broad question of 'banana's?' which... isn't really a question that has a clear answer.
In general it should perform better for exact questions in my epxerience

Dustin

Thibaultmol It's pretty much every query but for reference, this is "Command to prevent Mac from sleeping?"

This one has a clear answer but you can see the delays still exist there.

Thibaultmol

I think the answer is kinda the information you need right? the length in this example seems fine to me personally. Do you think it's too wordy?

And with delay, what exact delay do you mean? the fact that quick answer doesn't just appear in one go and instead has to stream in? (because that's how all LLM's work as you know)
(sorry, just trying to understand)

Dustin

Thibaultmol In this case, I feel it's too wordy for a "quick answer" as it gives way more information than is requested. The first sentence is enough. As a result, the entire process is lengthy and awkward as things slowly load in as the LLM generates the response. And additionally, 'Show More' takes several seconds to appear afterwards.

I do feel that the previous form of Quick Answer was much faster and concise.