Diogo Almeida says evals.typesafe.ai datasets will be easier to use
Diogo Almeida says that, nearly two weeks after launch and following requests, the datasets released with evals.typesafe.ai are being made easier to work with.
He says the change is in the spirit of openness, while he remains opposed to public benchmarks. On that basis, the team deprecates every dataset it evaluates on, whether the evaluation is internal or public.