List your company

Jailbreak Evaluation on Harmful Prompts Curated April 13, 2023

0Bad Bot RateAIM
-4235077Jul 5, 2023

Evaluation results

MethodLinks
AIMTarget Model=Claude v1.32023.0701000
few_shot_jsonTarget Model=Claude v1.32023.0701000
dev_mode_v2Target Model=Claude v1.32023.0701000
wikipedia_with_titleTarget Model=Claude v1.32023.0701000
wikipediaTarget Model=Claude v1.32023.0701000
prefix_injectionTarget Model=Claude v1.32023.0701000
distractors_negatedTarget Model=Claude v1.32023.0701000
base64_input_onlyTarget Model=Claude v1.32023.070973
leetspeakTarget Model=Claude v1.32023.0701000
prefix_injection_helloTarget Model=Claude v1.32023.0701000
noneTarget Model=Claude v1.32023.0701000
refusal_suppression_invTarget Model=GPT-42023.070973
refusal_suppression_invTarget Model=Claude v1.32023.0701000
evil_confidantTarget Model=GPT-42023.0701000
evil_confidantTarget Model=Claude v1.32023.0701000
rot13Target Model=Claude v1.32023.073691
base64_rawTarget Model=Claude v1.32023.073943
base64_output_onlyTarget Model=Claude v1.32023.073943
noneTarget Model=GPT-42023.073943
disemvowelTarget Model=Claude v1.32023.076913
base64_output_onlyTarget Model=GPT-42023.076940
prefix_injection_helloTarget Model=GPT-42023.076913
dev_mode_with_rantTarget Model=Claude v1.32023.079910
style_injection_jsonTarget Model=Claude v1.32023.079910
base64_input_onlyTarget Model=GPT-42023.079883
leetspeakTarget Model=GPT-42023.079846
auto_obfuscationTarget Model=Claude v1.32023.0712789
poemsTarget Model=GPT-42023.0712880
poemsTarget Model=Claude v1.32023.0712880
refusal_suppressionTarget Model=Claude v1.32023.0716840
disemvowelTarget Model=GPT-42023.0716813
rot13Target Model=GPT-42023.07162262
base64_rawTarget Model=GPT-42023.0716813
distractors_negatedTarget Model=GPT-42023.0719810
style_injection_shortTarget Model=GPT-42023.0722780
auto_obfuscationTarget Model=GPT-42023.0722699
prefix_injectionTarget Model=GPT-42023.0722780
style_injection_jsonTarget Model=GPT-42023.0724696
style_injection_shortTarget Model=Claude v1.32023.0725750
refusal_suppressionTarget Model=GPT-42023.0725723
auto_payload_splittingTarget Model=GPT-42023.07343828
base64Target Model=GPT-42023.0734660
base64Target Model=Claude v1.32023.0738566
wikipediaTarget Model=GPT-42023.07384716
distractorsTarget Model=GPT-42023.0744506
distractorsTarget Model=Claude v1.32023.0747530
dev_mode_with_rantTarget Model=GPT-42023.0750473
wikipedia_with_titleTarget Model=GPT-42023.07503119
evil_system_promptTarget Model=GPT-42023.0753470
few_shot_jsonTarget Model=GPT-42023.0753416
dev_mode_v2Target Model=GPT-42023.0753443
combination_1Target Model=GPT-42023.0756349
auto_payload_splittingTarget Model=Claude v1.32023.07592516
combination_1Target Model=Claude v1.32023.07661916
combination_2Target Model=GPT-42023.07691219
AIMTarget Model=GPT-42023.0775196
combination_3Target Model=Claude v1.32023.0781612
combination_2Target Model=Claude v1.32023.0784016
combination_3Target Model=GPT-42023.079433
Adaptive attackTarget Model=GPT-42023.071000—
Adaptive attackTarget Model=Claude v1.32023.071000—