Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Safety Detection on Full (test)
Loading...
64.3
TPR @ 3% FPR
Geometry-Lite
45.892
50.671
55.45
60.229
May 18, 2026
TPR @ 3% FPR
TPR @ 5% FPR
Updated 13d ago
Evaluation Results
Method
Method
Links
TPR @ 3% FPR
TPR @ 5% FPR
Geometry-Lite
Backbone Aggregation=P...
2026.05
64.3
75
MultiLayer-Linear
Backbone Aggregation=P...
2026.05
64.1
74.7
TaT-Disp-LSTM
Backbone Aggregation=P...
2026.05
56.7
69.4
Best single-layer probe
Backbone Aggregation=P...
2026.05
56.4
68.6
MultiLayer-DIM
Backbone Aggregation=P...
2026.05
46.6
57.4
Feedback
Search any
task
Search any
task