[xyzsparetimexyz]

Couldn't someone build an adaptive system, where the llm is frequently judging the difficulty of a task and switching to a more/less powerful model?

[iamflimflam1]

This only works well if you are creating sub agents with clean contexts for each task. If you constantly are switching models part way through some work then the whole session needs to be replayed each switch. You lose all the benefits of the context cache.

[xyzsparetimexyz]

Well yeah, it'd require the models to be loaded on the same system and the cache to be shared between them somehow

[inigyou]

Cache sharing is not possible. The numbers in the cache are completely specific to the model.

[xyzsparetimexyz]

Currently. I'm sure that you could make a system where the cache values are a superset C of e.g. models A and B where C is probably bigger than max(A,B) but smaller than A+B

[wonnage]

Reddit post so take with a grain of salt but this does seem possible. But unclear whether this is actually a viable architecture https://www.reddit.com/r/LocalLLaMA/comments/1t8s83r/nvidia_...

[inigyou]

It's equal to A+B. There is literally no sharing possible.

[xyzsparetimexyz]

Even when training both models together in a novel way?

[bob1029]

The judgment of task difficulty is a cursed problem. In most domains you really don't know what is down there until you start digging a little bit.

[xyzsparetimexyz]

Well yeah. You start on as medium model, it does the digging says says 'this task is too easy/hard/just right for me' and go from there.

[bkotrys]

definitely something that will come sooner than later. In my view, adaption of model routing / model orchestration and mixing providers is just natural next step.

[zombot]

And while you're at it, solve the halting problem.