That's really surprising! It kind of implies that LLMs are an incredibly effective lossy compression of the training inputs. I would not have thought that 3B weights would be enough to memorize texts ...
In honor of Naperville Public Library’s July celebration of “The Party That Shall Not Be Named” and all things Harry Potter, library staff were given the task of ranking each of J.K. Rowling’s seven ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results