INDEX

Explanations

winning by a landslide

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 peł

-0.95

 endlessly

-0.84

 incorrectly

-0.83

cially

-0.82

 actively

-0.82

 etkili

-0.81

pagar

-0.81

ually

-0.80

 pouvez

-0.79

optimal

-0.78

POSITIVE LOGITS

by

2.09

 convincingly

1.67

 comfortably

1.55

 handsomely

1.52

 soundly

1.41

 comfortable

1.41

 convincing

1.40

 easily

1.37

 convin

1.28

 easy

1.27

Activations Density 0.044%