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

CATQA

Benchmarks

Task NameDataset NameSOTA ResultTrend
Red-teamingCatQA
ASR0
20
Safety EvaluationCATQA FigStep
Safety Score100
8
Safety EvaluationCATQA Direct
Safety Score (1-ASR)1
8
Unsafety evaluationCATQA
Chinese Safety Score37.7
8
Jailbreak AttackCatQA-en
Harmful Rate1.01
6
Safety evaluationCATQA Chinese (test)
Chinese Accuracy55.67
4
Safety evaluationCATQA Vietnamese (test)
Score (Chinese)0.5336
4
Safety EvaluationCatQA-en (test)
Harmful Content Rate1.66
3
Showing 8 of 8 rows