3 ms·Bullshit benchmark for LLMs1 points by gpvos 8mo agonoemit 8mo agoThe underlying data looks scarce. If there's only a few questions per "category" of bullshit they can easily be gamed to favor one model over another.