X · @teortaxesTex
· X / Twitter
RT Jenia Jitsev 🏳️🌈 🇺🇦 🇮🇱 🇮🇷: Art of benchmarking 2: lower the eval scores of the reference model you compare to such that …
RT Jenia Jitsev 🏳️🌈 🇺🇦 🇮🇱 🇮🇷Art of benchmarking 2: lower the eval scores of the reference model you compare to such that your model appears "frontier-level champion". SOOFI S trained again the already fully open Nemotron-3-Nano. Report shows for Nemotron-3-Nano strongly lower scores than the original report