Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Counterfactual

Benchmarks

Task NameDataset NameSOTA ResultTrend
Knowledge cutoff successCounterfactual
Hard Unlearn Success Rate71.2
21
Compositional Out-of-Distribution DetectionCounterfactual dataset
AUC0.942
17
Counterfactual error detectionCounterfactual dataset
Finding Detection Rate24
5
Stereotyping EvaluationCounterfactual Open-Ended (OCF)
Stereotype Fraction2.5
5
Showing 4 of 4 rows