Comment by jttnr

I wonder, given Simons reputation in AI benchmarking, whether model providers try to train or tweak their models to perform better at drawing bicycles and pelicans?