List your company

Adversarial Attack Success on 12 Adversarial Tasks

72.7Attack Success RateEcho Chamber
-2.90816.72136.3555.979Jan 9, 2026

Evaluation results

MethodLinks
Echo ChamberModel=Gemini 2.5 Flash2026.0172.7
Echo ChamberModel=Gemini 2.0 Flash2026.0158.3
Echo ChamberModel=GPT-4.1 mini2026.0154.5
CrescendoModel=Gemini 2.0 Flash2026.0150
Echo ChamberModel=GPT-4.12026.0150
Echo ChamberModel=Gemini 2.5 Pro2026.0145.5
CrescendoModel=Gemini 2.5 Flash2026.0133.3
CrescendoModel=Gemini 2.5 Pro2026.0133.3
DANModel=Gemini 2.5 Flash2026.0125
DANModel=Gemini 2.0 Flash2026.0125
CrescendoModel=GPT-4.1 mini2026.0125
CrescendoModel=GPT-5 nano2026.0125
Echo ChamberModel=GPT-5 mini2026.0118.2
CrescendoModel=GPT-4.12026.0116.7
DANModel=Gemini 2.5 Pro2026.0116.7
CrescendoModel=GPT-5 mini2026.0116.7
Echo ChamberModel=GPT-5 nano2026.0116.7
DANModel=GPT-4.1 mini2026.010
DANModel=GPT-4.12026.010
DANModel=GPT-5 mini2026.010
DANModel=GPT-5 nano2026.010