INDEX

Explanations

recent decades ago

violated a rule

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

Entrepreneur

0.45

￡

0.45

serrat

0.45

Reliance

0.42

Communication

0.42

Über

0.42

 Prins

0.42

Kont

0.41

Bild

0.41

Após

0.41

POSITIVE LOGITS

 violate

0.55

 Росії

0.53

 violated

0.51

 росій

0.51

 violates

0.50

 clarified

0.49

 trochu

0.49

 エネルギー

0.48

 condiments

0.47

 violation

0.47

Activations Density 0.002%