This could work for enterprise solutions in companies if the prompt and order of data would not change much for operations. Suppose the language model is orchestrating something in the middle.
Tokenisation part is interesting. The choice of tokenisation can make or break a model especially for languages other than English.
Permutation invariance is the biggest drawback though.
This could work for enterprise solutions in companies if the prompt and order of data would not change much for operations. Suppose the language model is orchestrating something in the middle.
Tokenisation part is interesting. The choice of tokenisation can make or break a model especially for languages other than English.
Permutation invariance is the biggest drawback though.
Looking forward to what they come up with.😃