Hallucination rates are vanity. A model may pass one test but hit 30% errors on...

https://josuerwqt559.bearsfanteamshop.com/why-do-content-errors-stay-high-even-after-web-search-83-9-to-29-5

Hallucination rates are vanity. A model may pass one test but hit 30% errors on HalluHard. Because Vectara’s HHEM and AA-Omniscience measure facts differently, your yardstick is key