List your company

Healthcare Record Management on EHRAgent

74.8Accuracy (ACC)AGENTPOISON
19.1633.60548.0562.495Jul 17, 2024

Evaluation results

MethodLinks
AGENTPOISONAgent Backbone=ChatGPT, Retriever Type=end-to-end-retriever2024.0774.897.991.753.7
Non-attackAgent Backbone=ChatGPT, Retriever Type=contrastive-retriever2024.0773———
AGENTPOISONAgent Backbone=ChatGPT, Retriever Type=contrastive-retriever2024.0772.998.997.958.3
BadChainAgent Backbone=ChatGPT, Retriever Type=end-to-end-retriever2024.0771.921.150.533.7
Non-attackAgent Backbone=ChatGPT, Retriever Type=end-to-end-retriever2024.0771.6———
BadChainAgent Backbone=ChatGPT, Retriever Type=contrastive-retriever2024.0770.811.272.58.3
GCGAgent Backbone=ChatGPT, Retriever Type=end-to-end-retriever2024.0770.812.563.530.2
GCGAgent Backbone=ChatGPT, Retriever Type=contrastive-retriever2024.0770.19.481.345.8
AutoDANAgent Backbone=ChatGPT, Retriever Type=contrastive-retriever2024.0768.484.289.527.4
CPAAgent Backbone=ChatGPT, Retriever Type=contrastive-retriever2024.0767.996.958.351.1
AutoDANAgent Backbone=ChatGPT, Retriever Type=end-to-end-retriever2024.0767.438.951.642.1
CPAAgent Backbone=ChatGPT, Retriever Type=end-to-end-retriever2024.0766.361.555.838.5
GCGAgent Backbone=LLaMA3, Retriever Type=contrastive-retriever2024.0744.216.414.829.5
AGENTPOISONAgent Backbone=LLaMA3, Retriever Type=contrastive-retriever2024.074110021.565.6
Non-attackAgent Backbone=LLaMA3, Retriever Type=contrastive-retriever2024.0737.7———
AutoDANAgent Backbone=LLaMA3, Retriever Type=contrastive-retriever2024.0736.175.46.657.4
CPAAgent Backbone=LLaMA3, Retriever Type=contrastive-retriever2024.0734.496.924.672.1
GCGAgent Backbone=LLaMA3, Retriever Type=end-to-end-retriever2024.0734.48.811.519.7
AGENTPOISONAgent Backbone=LLaMA3, Retriever Type=end-to-end-retriever2024.0734.496.77.768.9
Non-attackAgent Backbone=LLaMA3, Retriever Type=end-to-end-retriever2024.0732.8———
AutoDANAgent Backbone=LLaMA3, Retriever Type=end-to-end-retriever2024.0731.113.11.68.2
BadChainAgent Backbone=LLaMA3, Retriever Type=end-to-end-retriever2024.0731.16.28.213.1
BadChainAgent Backbone=LLaMA3, Retriever Type=contrastive-retriever2024.0729.531.11865.6
CPAAgent Backbone=LLaMA3, Retriever Type=end-to-end-retriever2024.0721.315.34.88.6