Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Privacy and Utility Evaluation on AgentLeak Hierarchical (5 agents)
Loading...
85
Privacy
Full-System LCGuard
22.6
38.8
55
71.2
May 21, 2026
Privacy
Task Performance
Helpfulness
Leakage
ASR
Updated 12d ago
Evaluation Results
Method
Method
Links
Privacy
Task Performance
Helpfulness
Leakage
ASR
Full-System LCGuard
Backbone=Gemma-9B, Top...
2026.05
85
63
70.5
15
21.5
ADAPT
Backbone=Gemma-9B, Top...
2026.05
82.2
33
27
17.8
38
Per-Agent LCGuard
Backbone=Gemma-9B, Top...
2026.05
78.6
60
64
21.5
25.4
PrivAct
Backbone=Gemma-9B, Top...
2026.05
56
58
39
44
74
Vanilla KV Sharing (LatentMAS)
Backbone=Gemma-9B, Top...
2026.05
25
65
76.5
75
90
Feedback
Search any
task
Search any
task