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

Jailbreak Defense on GCG

0ASR

PPL

Updated 1mo ago

Evaluation Results

Method
PPL 2024.02	0	1.02
PPL 2024.02	0	1.06
Self-Reminder 2024.02	0	1
ICD 2024.02	0	1
SafeDecoding 2024.02	0	1
Self-Examination 2025.05	0	1.08
Goal Prioritization 2025.05	0	1
IA 2025.05	0	1
SAGE 2025.05	0	1
Self-Examination 2025.05	0	1
Goal Prioritization 2025.05	0	1
SAGE 2025.05	0	1
ICD 2025.05	0	1
SAGE 2025.05	0	1
PreSafe 2026.03	0	-
R2D 2026.03	0	-
PreSafe 2026.03	0	-
R2D 2026.03	0	-
R2D 2026.03	1	-
R2D 2026.03	1	-
PreSafe 2026.03	1	-
Retokenization 2024.02	2	1
Self-Reminder 2025.05	2	1
IA 2025.05	2	1.08
Perplexity Filter 2026.03	3	-
PreSafe 2026.03	3	-
SafeDecoding 2024.02	4	1.12
Paraphrase 2024.02	4	1.06
Self-Reminder 2025.05	4	1.1
Erase-And-Check 2026.03	4	-
PreSafe 2026.03	4	-
ICD 2025.05	6	1.24
ICD 2025.05	6	1.18
Self-Examination 2025.05	6	1.06
Erase-And-Check 2026.03	6.1	-
PreSafe 2026.03	6.1	-
Perplexity Filter 2026.03	6.1	-
Erase-And-Check 2026.03	6.1	-
R2D 2026.03	6.1	-
IA 2025.05	7	1.18
Erase-And-Check 2026.03	7.1	-
Erase-And-Check 2026.03	8.1	-
Perplexity Filter 2026.03	9.1	-
Erase-And-Check 2026.03	9.1	-
Perplexity Filter 2026.03	9.1	-
Self-Examination 2024.02	12	1.4
Self-Examination 2024.02	12	1.56
Self-Reminder 2025.05	12	1.24
Perplexity Filter 2026.03	12.1	-
Smooth LLM 2026.03	14.1	-
Perplexity Filter 2026.03	17.2	-
R2D 2026.03	17.2	-
Smooth LLM 2026.03	18.2	-
SafeChain 2026.03	19.2	-
Paraphrase 2024.02	20	1.8
Smooth LLM 2026.03	21.2	-
No Defense 2025.05	22	1.88
SafeChain 2026.03	23.2	-
Remove Non-Dictionary 2026.03	25.3	-
None 2026.03	26.2	-
Smooth LLM 2026.03	26.3	-
None 2026.03	27.2	-
SafeChain 2026.03	27.3	-
Smooth LLM 2026.03	27.3	-
Goal Prioritization 2025.05	28	1.36
Remove Non-Dictionary 2026.03	28.3	-
None 2026.03	30.3	-
No Defense 2024.02	32	2.48
SafeChain 2026.03	32.3	-
Remove Non-Dictionary 2026.03	32.3	-
Synonym Substitution 2026.03	33.3	-
Synonym Substitution 2026.03	34.3	-
SafeChain 2026.03	34.3	-
SafeChain 2026.03	35.4	-
None 2026.03	37.4	-
Synonym Substitution 2026.03	37.4	-
Remove Non-Dictionary 2026.03	39.4	-
Smooth LLM 2026.03	39.4	-
Retokenization 2024.02	42	1.58
Self-Reminder 2024.02	42	2.76
Synonym Substitution 2026.03	44.4	-
None 2026.03	48.5	-
Remove Non-Dictionary 2026.03	48.5	-
Synonym Substitution 2026.03	50.5	-
Synonym Substitution 2026.03	51.5	-
None 2026.03	52.5	-
Remove Non-Dictionary 2026.03	55.6	-
No Defense 2025.05	60	3
ICD 2024.02	70	3.86
No Defense 2025.05	96	4.74
No Defense 2024.02	100	4.7