Introduces ETHICS benchmark showing current language models have promising but incomplete ability to predict basic human ethical judgments on text scenarios.
For Justice, Deontology, and Commonsense Morality, we had 5 MTurkers relabel each example, and we kept examples for which at least 4 out of the 5 agreed
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CY 1years
2020 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Aligning AI With Shared Human Values
Introduces ETHICS benchmark showing current language models have promising but incomplete ability to predict basic human ethical judgments on text scenarios.