atyh on nostr

35 AI models tested across 172 billion tokens of real document questions.

median hallucination rate:
25%.
One out of every 4 answers contain false information.

53❤️6👀1🤯1

Ape Mithrandir · 5w

But we out here using it for target selection in WW3.

Cat-Go-Purrrrrrr · 5w

my stoner hippie friend has a lower average hallucination score

Abstract Equilibrium · 5w

Now you read all those documents and when I quiz you, I challenge you to do better! 🤣