Ring-a-bell
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Ring-A-Bell Nudity
95.6Attack Success Rate (ASR)
52
Ring-A-Bell
98ASR
23
Ring-A-Bell Violence
88Attack Success Rate (ASR)
22
Ring-A-Bell Violence (test)
5.54ASR
21
Ring-A-Bell Nudity (test)
9.22ASR
21
Ring-A-Bell
0.83Ring-A-Bell Score
20
Ring-A-Bell
1.3ASR
18
Ring-A-Bell
98.2Generation Rate
17
Ring-A-Bell (test)
0Attack Success Rate
15
Ring-A-Bell 77
92.6Attack Success Rate (ASR)
15
Ring-A-Bell K-16, K-38, K-77
0.958K-16 Score
14
Ring-A-Bell 1.5 (test)
17.9Nudity Generation Rate
13
Ring-A-Bell
93ASR
13
Ring-A-Bell
4.93Ring-16 Score
13
Ring-A-Bell
83.1ASR
13
Ring-a-bell Adversarial Prompts
100Erase Rate (%)
13
Ring-A-Bell 5-77
99.2ASR
12
Ring-A-Bell adversarial prompts (K77)
1.05ASR (Threshold 0.3)
10
Ring-A-Bell adversarial prompts K38
1.05ASR (T=0.3)
10
Ring-A-Bell adversarial prompts (K16)
71.58ASR (T=0.3)
10
Ring-A-Bell
5.1Attack Success Rate (ASR)
9
Ring-A-Bell Union
24.77Nudity Rate
9
Ring-A-Bell
11.01Nudity Rate
9
Ring-A-Bell (285 prompts)
44.56ASR
8
Ring-a-bell (adversarial)
100Q16 IP
8