Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Prefill Attack Defense on ClearHarm prefill (held-out)
Loading...
52
PAR
Base model
33.28
38.14
43
47.86
Jun 4, 2026
PAR
Updated 1mo ago
Evaluation Results
Method
Method
Links
PAR
Base model
Note=no training
2026.06
52
BCT
Train threat=Frustration
2026.06
40
MLPCT
Train threat=Sycophancy
2026.06
34
BCT
Train threat=Jailbreak
2026.06
34
Feedback
Search any
task
Search any
task
