Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

AgentLeak

Benchmarks

Task NameDataset NameSOTA ResultTrend
Multi-agent Sequential CommunicationAgentLeak
Privacy Score86.5
20
Communication channel leakage evaluationAgentLeak (test)
Privacy Score0.823
10
Multi-agent privacy and utility evaluationAgentLeak Hierarchical
Privacy87
10
Multi-agent privacy and utility evaluationAgentLeak Sequential
Privacy86
10
Privacy and Utility EvaluationAgentLeak Hierarchical 5 agents
Privacy85
5
Privacy and Utility EvaluationAgentLeak Sequential, 3 agents
Privacy Score84.5
5
Multi-agent latent communication privacy and utilityAgentLeak Graph
Privacy78
5
Privacy Leakage AnalysisAgentLeak traces
C1 Output27.2
3
Showing 8 of 8 rows