AI Models Can’t Agree on Basic Facts Most of the Time, Study Shows
In brief Five frontier AI models disagreed on 67% of 1,000 real-world fact-check claims. Unanimous agreement happened on only 328 claims. At 0.639 Krippendorff’s alpha, the models fall below the 0.8 reliability threshold. Ask five […]
