List your company

Task-specific Risk Detection on EICU-AC (test)

98.7LPAGuardAgent
46.760.273.787.2Feb 17, 2025

Evaluation results

MethodLinks
GuardAgentType=Guardrail-based, Backbone=GPT-42025.0298.710097.598.798.7
AGrailType=Guardrail-based, Backbone=Claude-3.52025.0298.49710098.5100
Claude-3.5Type=Model-based, Shots=0-shot2025.0297.995.910097.9100
AGrailType=Guardrail-based, Backbone=GPT-4o2025.0297.897.598.197.8100
GPT-4oType=Model-based, Shots=0-shot2025.0297.294.710097.3100
Claude-3.5Type=Model-based, Shots=1-shot2025.0294.695.394.494.7100
GPT-4oType=Model-based, Shots=1-shot2025.0292.189.695.792.5100
AgentMonitorType=Guardrail-based, Backbone=GPT-4o2025.0282.398.266.779.4100
LLaMA-Guard3Type=Guardrail-based2025.0248.7—0——