> An updated version achieved similar robustness on synthetic evaluations, and d...

		Vecr 11 days ago \| parent \| context \| favorite \| on: Constitutional Classifiers: Defending against univ... > An updated version achieved similar robustness on synthetic evaluations, and did so with a 0.38% increase in refusal rates and moderate additional compute costs. "Synthetic evaluations" aren't 70 hours of Pliny the Prompter.