Forget the hype about AI "solving" human cognition, new research suggests unified models like Centaur are just overfitted "black boxes" that fail to understand basic instructions.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results