INDEX

Explanations

Alan followed by names

refusing harmful requests

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 protection

0.43

 governing

0.39

 acces

0.39

იმ

0.38

 sensitive

0.38

_;

0.37

 coordinated

0.37

 PROTECTION

0.37

保護

0.37

 several

0.36

POSITIVE LOGITS

compassing

0.54

Alan

0.48

 Alan

0.46

brooke

0.43

 отлича

0.43

atore

0.41

 Turing

0.41

 alan

0.41

ah

0.39

 continuidade

0.39

Activations Density 0.006%