INDEX

Explanations

shop online

The neuron is keyed to “content” tokens—i.e. substantive identifiers and labels (names of UI strings, class/entity names, menu items, etc.), as opposed to punctuation or common function words.

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

䛗

-2.33

艄

-2.19

饜

-2.16

"../

-2.11

橚

-2.06

雠

-2.03

歺

-1.98

 этого

-1.94

 этом

-1.93

dicionado

-1.91

POSITIVE LOGITS

 menyadari

1.91

the

1.83

 Если

1.80

1.76

 важ

1.73

 tienen

1.66

is

1.65

1.62

 Even

1.59

可以

1.58

Activations Density 0.004%