Epoch AI argues good benchmarks should answer bigger questions than 'can the model do this one task,' and outlines what those are.
Continue to AI University →