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

Safety Evaluation on JailbreakBench (test)

1.12ASR (Llama-Guard-3-8B)

MLP

-2.170420.039842.2564.4602May 6, 2026
Updated 27d ago

Evaluation Results

MethodLinks
2026.05
1.124.51
2026.05
1.623.621.25
2026.05
4.122.752.62
2026.05
54.386.25
2026.05
5.014.625.75
2026.05
5.384.386
2026.05
10.387.3810.12
2026.05
11.19.6214.88
2026.05
11.129.8813.25
2026.05
13.121314
2026.05
149.2515.75
2026.05
151611.38
2026.05
1616.3814.75
2026.05
18.1215.7525.25
2026.05
18.3817.3816.38
2026.05
19.525.124.88
2026.05
21.75108.5
2026.05
27.3815.7517.38
2026.05
30.1620.2526.12
2026.05
30.2939.2528.25
2026.05
35.2534.3831.87
2026.05
35.882948
2026.05
37.525.1240.12
2026.05
39.3631.2542.75
2026.05
42.6254.8748.38
2026.05
44.538.6249.62
2026.05
45.1245.2548.38
2026.05
46.493955
2026.05
46.7540.2557.25
2026.05
49.555.8855.75
2026.05
49.6245.2553.12
2026.05
50.2549.3864
2026.05
52.3841.8857
2026.05
53.1439.7552.75
2026.05
55.2542.3852.25
2026.05
56.8254.3767.88
2026.05
59.5765.8858.5
2026.05
60.255080.5
2026.05
62.2553.6273.38
2026.05
63.0451.8889.88
2026.05
63.6252.548.25
2026.05
65.1247.7569
2026.05
65.8158.8186.38
2026.05
66.3753.4486.94
2026.05
66.8560.5690.44
2026.05
67.3857.3870.75
2026.05
68.1263.6261.12
2026.05
7267.3863.12
2026.05
73.095790.62
2026.05
73.3658.8181.69
2026.05
74.2553.8792.06
2026.05
75.8470.2590.12
2026.05
76.2561.2584
2026.05
78.3867.6285.5
2026.05
79.9763.1278.12
2026.05
83.386289.88