INDEX

Explanations

undercover investigations

acting undercover or secretly

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

uge

0.42

)(&

0.42

Iteration

0.42

kah

0.41

small

0.40

short

0.40

vskip

0.40

alach

0.39

tanh

0.39

 mutating

0.39

POSITIVE LOGITS

 undercover

0.61

 decoy

0.54

 entrap

0.50

 felony

0.49

 infiltr

0.47

 रिश्वत

0.47

 expectant

0.46

 infiltrate

0.44

 entendu

0.44

 entraî

0.43

Activations Density 0.034%