Why do LLMs make stuff up? New research peers under the hood.

JFC. Computers don't "make stuff up". They generate sentences based on statistics and programming. If it's "wrong", that's just the program. There's no inherent reason why these programs would be magic truth machines. Maybe these people need to figure out their own damn code but that's not some breakthrough. It's not some mystery that needs "research".

TL;DR: GIGO.