List your company

Jailbreaking on HarmBench 51 (test)

98.1ASR@5 (Standard)WILDTEAMING
49.7462.29574.8587.405Jun 26, 2024

Evaluation results

MethodLinks
WILDTEAMINGModel=Mixtral (8x7B)2024.0698.12.728.7587.28.990.7220.53155
PAIRModel=Mixtral (8x7B)2024.0697.53.059.5461.813.960.7950.53328
WILDTEAMINGModel=Tulu2 DPO (7B)2024.0696.92.618.7787.88.980.7220.52961
PAIRModel=Mistral (7B)2024.0695.63.289.626514.210.7920.53730
PAIRModel=Tulu2 DPO (7B)2024.06953.579.7862.114.240.7920.53429
WILDTEAMINGModel=Mistral (7B)2024.06952.378.5689.28.720.7220.52752
PAIRModel=Vicuna (7B)2024.0694.33.559.4259.514.780.790.5327
AUTODANModel=Tulu2 DPO (7B)2024.0694.3—12.97201.410.9720.96236
WILDTEAMINGModel=Vicuna (7B)2024.0693.12.828.6588.19.310.7220.52755
AUTODANModel=Mistral (7B)2024.0692.5—13.2419.9—0.9610.95240
WILDTEAMINGModel=GPT-3.5 (0613)2024.0692.57.087.9665.813.190.7330.52650
GCGModel=Vicuna (7B)2024.0689.9—4,062.57—————
AUTODANModel=Vicuna (7B)2024.0689.3—13.7419.4—0.9720.96936
AUTODANModel=Mixtral (8x7B)2024.0688.7—13.31201.530.9670.95738
PAIRModel=GPT-3.5 (0613)2024.0688.76.659.7861.217.010.7980.5326
GCGModel=Mistral (7B)2024.0685.5—2,266.69—————
WILDTEAMINGModel=GPT-4 (0613)2024.0679.98.618.1360.113.430.7310.5339
PAIRModel=GPT-4 (0613)2024.0678.69.649.3344.917.750.8020.53829
GCGModel=Tulu2 DPO (7B)2024.0651.6—4,265.86—————