← Back to context Comment by fragmede 5 hours ago 54.0% isn't particularly high. 1 comment fragmede Reply prodigycorp 2 hours ago What I'm trying to say which doesnt seem obvious is that all models are x% correct at benchmarks until they get saturated, and then new benchmarks get made.
prodigycorp 2 hours ago What I'm trying to say which doesnt seem obvious is that all models are x% correct at benchmarks until they get saturated, and then new benchmarks get made.
What I'm trying to say which doesnt seem obvious is that all models are x% correct at benchmarks until they get saturated, and then new benchmarks get made.