Jailbreaking on EHRAgent ALL
70.112Weighted Average ASRJailAgent
Evaluation Results
| Method | Links | |
|---|---|---|
| JailAgentAgent Backbone=GPT-4o2026.04 | 70.112 | |
| JailAgentAgent Backbone=Claude-3.5-haiku2026.04 | 69.336 | |
| JailAgentAgent Backbone=Gemini-3.0-pro2026.04 | 68.466 | |
| JailAgentAgent Backbone=GPT-52026.04 | 66.753 | |
| AgentPoisonAgent Backbone=GPT-4o2026.04 | 66.708 | |
| AgentPoisonAgent Backbone=Claude-3.5-haiku2026.04 | 65.763 | |
| BadChainAgent Backbone=Claude-3.5-haiku2026.04 | 63.639 | |
| BadChainAgent Backbone=GPT-4o2026.04 | 63.555 | |
| BadChainAgent Backbone=GPT-52026.04 | 62.619 | |
| AgentPoisonAgent Backbone=GPT-52026.04 | 62.142 | |
| JailAgentAgent Backbone=GPT-3.5-turbo2026.04 | 62.057 | |
| BadChainAgent Backbone=Gemini-3.0-pro2026.04 | 61.927 | |
| JailAgentAgent Backbone=Llama-3.1-70B2026.04 | 61.291 | |
| AgentPoisonAgent Backbone=Gemini-3.0-pro2026.04 | 61.15 | |
| PAIRAgent Backbone=Claude-3.5-haiku2026.04 | 58.559 | |
| AgentPoisonAgent Backbone=Llama-3.1-70B2026.04 | 58.54 | |
| BadChainAgent Backbone=Llama-3.1-70B2026.04 | 58.213 | |
| BadChainAgent Backbone=GPT-3.5-turbo2026.04 | 56.729 | |
| AgentPoisonAgent Backbone=GPT-3.5-turbo2026.04 | 56.136 | |
| PAIRAgent Backbone=GPT-52026.04 | 55.686 | |
| PAIRAgent Backbone=GPT-4o2026.04 | 54.379 | |
| PAIRAgent Backbone=Gemini-3.0-pro2026.04 | 53.854 | |
| PAIRAgent Backbone=Llama-3.1-70B2026.04 | 51.347 | |
| PAIRAgent Backbone=GPT-3.5-turbo2026.04 | 48.933 |