Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Model Stealing Defense on MedMCQA
Loading...
50.2
Agreement (%)
No Defense
42.504
44.502
46.5
48.498
Jun 14, 2026
Agreement (%)
Average Score
User Impact (%)
Updated 1mo ago
Evaluation Results
Method
Method
Links
Agreement (%)
Average Score
User Impact (%)
No Defense
Query budget (B)=500,...
2026.06
50.2
44
0
PRADA
Query budget (B)=500,...
2026.06
48.4
43.6
0.4
DRW
Query budget (B)=500,...
2026.06
48.1
43.1
4.8
HoneypotNet
Query budget (B)=500,...
2026.06
47.8
42.4
1.5
KT w/o Breadcrumb
Query budget (B)=500,...
2026.06
47.4
41.4
0
Output Pert.
Query budget (B)=500,...
2026.06
47.3
42.6
2.1
KT w/o Detection
Query budget (B)=500,...
2026.06
46.3
40.3
0
KT (Full)†
Query budget (B)=500,...
2026.06
42.8
37.8
0
Feedback
Search any
task
Search any
task