List your company

Red-teaming on 50 harmful goals (Manual evaluation)

100Hard ASRPAIR
-1.9224.545177.46Mar 3, 2026

Evaluation results

MethodLinks
PAIRTarget Model=Gemini 2.5 Flash2026.03100
ITMS-VDTarget Model=Qwen2.5-Omni2026.03100
PAIRTarget Model=GPT-4o2026.0398
ITMS-CrescendoTarget Model=Gemini 2.5 Flash2026.0398
CrescendoTarget Model=Gemini 3 Flash2026.0398
PAIRTarget Model=Qwen2.5-Omni2026.0398
CrescendoTarget Model=Qwen3-Omni2026.0398
CrescendoTarget Model=GPT-4o2026.0396
PAIRTarget Model=Gemini 3 Flash2026.0396
CrescendoTarget Model=Qwen2.5-Omni2026.0396
PAIRTarget Model=Qwen3-Omni2026.0396
CrescendoTarget Model=Gemini 2.5 Flash2026.0394
ITMS-CrescendoTarget Model=Gemini 3 Flash2026.0394
ITMS-CrescendoTarget Model=Qwen3-Omni2026.0394
ITMS-CrescendoTarget Model=Claude Sonnet 42026.0392
ITMS-CrescendoTarget Model=GPT-4o2026.0392
CrescendoTarget Model=Claude Sonnet 42026.0390
ITMS-CrescendoTarget Model=Qwen2.5-Omni2026.0388
Violent DurianTarget Model=Qwen2.5-Omni2026.0386
ITMS-VDTarget Model=Gemini 2.5 Flash2026.0362
PAIRTarget Model=Claude Sonnet 42026.0360
Violent DurianTarget Model=Gemini 2.5 Flash2026.0356
Violent DurianTarget Model=GPT-4o2026.0342
ITMS-VDTarget Model=GPT-4o2026.0340
ITMS-VDTarget Model=Gemini 3 Flash2026.0334
Violent DurianTarget Model=Qwen3-Omni2026.0330
Violent DurianTarget Model=Gemini 3 Flash2026.0326
ITMS-VDTarget Model=Qwen3-Omni2026.0322
ITMS-VDTarget Model=Claude Sonnet 42026.036
Violent DurianTarget Model=Claude Sonnet 42026.032