INDEX

Explanations

error analysis

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 situation

-0.69

 contribution

-0.67

writeFieldEnd

-0.65

 perſon

-0.65

 issue

-0.65

ials

-0.64

 defaultstate

-0.63

 Gegenteil

-0.62

 houſe

-0.62

Efq

-0.62

POSITIVE LOGITS

are

0.57

orianCalendar

0.56

 cherchés

0.55

GEBURTSDATUM

0.54

 zostały

0.53

 validamos

0.52

są

0.51

 aren

0.51

 restent

0.51

are

0.50

Activations Density 0.026%