Originalarbeiterreichbar
GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Tasks
Benchmark von Open
geprüft 10.11.2026
Worauf sich diese Seite beruft, wörtlich, abgerufen am 11.10.2026:
GDPval covers the majority of U.S. Bureau of Labor Statistics Work Activities for 44 occupations
bestätigt 10.11.2026Tasks are constructed from the representative work of industry professionals with an average of 14 years of experience
bestätigt 10.11.2026is an estimate of the time taken to assess a model deliverable by a human expert grader
bestätigt 10.11.2026is how often the model deliverable is rated better than the human deliverable by the human expert grader
bestätigt 10.11.2026then does the task themselves if the deliverable quality is below their quality bar
bestätigt 10.11.2026When incorporating time to review and redo work, the payoff from using a model shrinks
bestätigt 10.11.2026We do not include consideration of the time taken to review a human professional deliverable
bestätigt 10.11.2026it does not capture the cost of catastrophic mistakes
bestätigt 10.11.2026