INDEX

Explanations

numbers and codes

The neuron lights up on salient content tokens—i.e. the main words of a text (nouns, verbs, numbers, etc.)—rather than on purely grammatical “filler” tokens.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 Francis

-1.00

箕

-0.88

):

-0.86

 remar

-0.84

horas

-0.84

Dun

-0.83

 peix

-0.82

脈

-0.82

 scor

-0.81

 obok

-0.80

POSITIVE LOGITS

 kän

0.94

Though

0.91

 vilja

0.88

 gravidanza

0.85

 berhubungan

0.82

 बताया

0.79

kve

0.78

Só

0.77

提起

0.77

diven

0.77

Activations Density 0.191%