Panel Presentation: Liquid Neural Networks with Live Q&A

[@MonkkSoori]

1) In the slide at 28:50 you say that the liquid time-constant differential equation is solved in closed form (which I assume means it is an exact solution) and you provide the solution expression , but then you say in 31:00 that this same expression is an approximation of the closed-form solution. How can that be? 2) And then in 32:30, it is not at all mentioned how the new x(t) expression is attained and related to the closed-form solution at the bottom of the slide. Please explain.

[@minditon3264]

You are really doing your work I guess

[@ps3301]

Where is the multimodal liquid neural network that can do inference for vision, language and sound ?

[@adiexplains]

Why isnt this become mainstream? I also see that your Liquid 40B model is not performing as it is claimed. It is not able to perform very simple language tasks. I also suspect the benchmark that you have put in your website.