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

Safety Alignment on Harmful Dataset (test)

81Harmful Score

Non-Aligned

32.32844.96457.670.236Feb 2, 2024
Updated 1mo ago

Evaluation Results

MethodLinks
2024.02
81
2024.02
79.2
2024.02
77.6
2024.02
65.6
2024.02
60
2024.02
56.6
2024.02
56.6
2024.02
55.2
2024.02
55.2
2024.02
54.4
2024.02
54.4
2024.02
54
2024.02
54
2024.02
53.6
2024.02
53.6
2024.02
52.6
2024.02
50.6
2024.02
50.6
2024.02
50.6
2024.02
50.6
2024.02
50.6
2024.02
50
2024.02
49.8
2024.02
49.4
2024.02
48.6
2024.02
48.2
2024.02
42.8
2024.02
42.4
2024.02
42.2
2024.02
34.2