List your company

Jailbreak evaluation on curated dataset (test)

0BAD BOT Ratebase64_input_only
-0.040.230.50.77Jul 5, 2023

Evaluation results

MethodLinks
base64_input_onlyModel=GPT-3.5 Turbo, Encoding=Base64 Input2023.0700.530.47
base64_output_onlyModel=GPT-3.5 Turbo, Encoding=Base64 Output2023.0700.090.91
base64_rawModel=GPT-3.5 Turbo, Encoding=Base64 Raw2023.07001
noneModel=GPT-3.5 Turbo, Context=Control group2023.070.030.970
base64Model=GPT-3.5 Turbo, Encoding=Base642023.070.030.060.91
style_injection_shortModel=GPT-3.5 Turbo2023.070.690.190.12
dev_mode_v2Model=GPT-3.5 Turbo, Attack Source=jailbreakchat.com2023.070.780.220
evil_system_promptModel=GPT-3.5 Turbo, Attack Source=jailbreakchat.com2023.070.880.090.03
AIMModel=GPT-3.5 Turbo, Attack Source=jailbreakchat.com2023.070.970.030
dev_mode_with_rantModel=GPT-3.5 Turbo, Attack Source=jailbreakchat.com2023.070.970.030
Adaptive attackModel=GPT-3.5 Turbo2023.0710—