What's amazing is that given how LLMs receive input data (as tokenized streams, as other commenters have pointed out) it's remarkable that it can ever answer th

What's amazing is that given how LLMs receive input data (as tokenized streams, as other commenters have pointed out) it's remarkable that it can ever answer this question correctly.