The real improvement will be when the software engineers get into the training loop. Then we can have MoE that use cache-friendly expert utilisation and maybe even learned prefetching for what the next experts will be.
reply> maybe even learned prefetching for what the next experts will be
replyExperts are predicted by layer and the individual layer reads are quite small, so this is not really feasible. There's just not enough information to guide a prefetch.
It's feasible to put the expert routing logic in a previous layer. People have done it: https://arxiv.org/abs/2507.20984
replyManually no. It would have to be learned, and making the expert selection predictable would need to be a training metric to minimize.
replyMaking the expert selection more predictable also means making it less effective. There's no real free lunch.
reply