Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Safety Detection on JBB-Behaviors (held-out)
Loading...
93.7
AUROC
MultiLayer-DIM
82.78
85.615
88.45
91.285
May 18, 2026
AUROC
Updated 13d ago
Evaluation Results
Method
Method
Links
AUROC
MultiLayer-DIM
Protocol=Leave-one-ben...
2026.05
93.7
TaT-Disp-LSTM
Protocol=Leave-one-ben...
2026.05
87.9
MultiLayer-Linear
Protocol=Leave-one-ben...
2026.05
87.1
Geometry-Lite
Protocol=Leave-one-ben...
2026.05
86.2
Best single-layer probe
Protocol=Leave-one-ben...
2026.05
83.2
Feedback
Search any
task
Search any
task