3 ms·
This seems like it could be a game changer. Modern LLM based applications face a balancing act of context limitations, which often results in some kind of mapre
by helen___keller 3y ago
This seems like it could be a game changer. Modern LLM based applications face a balancing act of context limitations, which often results in some kind of mapreduce-type behavior when that context can’t fit the input
If contexts keep growing, the landscape of LLM application engineering will as well
- whimsicalism 3y agoThe problem is there are no public benchmarks usually so it is hard to really compare on long context lengths to see if they are still performing equally intelligent of tasks.