GDPval-AA v2 Benchmark Validity Questioned Despite High Weighting

GDPval-AA v2 uses frontier models to judge other AI outputs on questions derived from a closed benchmark, then weights that heavily in the Intelligence Index v4.1. The human ELO baseline methodology is unspecified, making the 1818 ELO score for Claude Fable 5 difficult to interpret against real-world capability.