List your company

Jailbreaking on EHRAgent MIMIC-III

56.55SRNon-attack
33.066839.163445.2651.3566Apr 7, 2026

Evaluation results

MethodLinks
Non-attackAgent Backbone=GPT-52026.0456.55———87.76
Non-attackAgent Backbone=GPT-4o2026.0455.86———84.31
JailAgentAgent Backbone=GPT-4o2026.0455.5265.1759.4856.2183.62
JailAgentAgent Backbone=GPT-52026.0455.5266.3860.1756.7286.72
AgentPoisonAgent Backbone=GPT-52026.0452.0762.4158.6254.1487.93
AgentPoisonAgent Backbone=GPT-4o2026.0451.963.2851.2151.0380.34
BadChainAgent Backbone=GPT-52026.0451.3862.0758.9757.5986.03
BadChainAgent Backbone=GPT-4o2026.0447.7662.7649.6648.178.62
PAIRAgent Backbone=GPT-52026.0447.7651.947.2444.4884.14
JailAgentAgent Backbone=Gemini-3.0-pro2026.0447.4148.9744.8345.3488.45
Non-attackAgent Backbone=GPT-3.5-turbo2026.0446.72———56.9
Non-attackAgent Backbone=Gemini-3.0-pro2026.0446.72———89.48
AgentPoisonAgent Backbone=Claude-3.5-haiku2026.0446.3870.5266.965.5297.93
Non-attackAgent Backbone=Claude-3.5-haiku2026.0446.03———98.62
JailAgentAgent Backbone=GPT-3.5-turbo2026.0445.8664.3159.3155.8656.03
JailAgentAgent Backbone=Claude-3.5-haiku2026.0445.1770.3467.076597.59
BadChainAgent Backbone=Claude-3.5-haiku2026.0444.1467.4165.1762.7696.21
AgentPoisonAgent Backbone=Gemini-3.0-pro2026.0443.2845.3442.4141.5586.03
JailAgentAgent Backbone=Llama-3.1-70B2026.0442.9363.9761.2158.9781.9
Non-attackAgent Backbone=Llama-3.1-70B2026.0442.41———81.38
PAIRAgent Backbone=Claude-3.5-haiku2026.0442.0757.9351.7249.4894.31
BadChainAgent Backbone=Gemini-3.0-pro2026.0442.0743.9740.5240.1782.07
BadChainAgent Backbone=Llama-3.1-70B2026.0441.7262.4157.0756.2178.62
AgentPoisonAgent Backbone=Llama-3.1-70B2026.0441.5561.956.3855.8678.45
PAIRAgent Backbone=GPT-4o2026.0440.524538.2838.2873.28
BadChainAgent Backbone=GPT-3.5-turbo2026.0439.8361.0354.3151.2143.97
PAIRAgent Backbone=Gemini-3.0-pro2026.0439.8336.0334.6634.3178.62
AgentPoisonAgent Backbone=GPT-3.5-turbo2026.0438.2858.7951.3849.4842.41
PAIRAgent Backbone=GPT-3.5-turbo2026.0435.8642.2437.0736.5539.31
PAIRAgent Backbone=Llama-3.1-70B2026.0433.9752.4148.147.7670.17