INDEX

Explanations

nouns or concepts

The neuron fires strongly on content-bearing nouns and key technical terms (e.g. “outcome,” “dialogue,” “illumination,” “movement,” “performance,” “answers”) rather than on 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

:]:

-0.97

锖

-0.96

}:=\

-0.96

 élève

-0.95

such

-0.93

vää

-0.93

Ꮭ

-0.93

 şeyler

-0.91

Penulis

-0.90

 включая

-0.88

POSITIVE LOGITS

 that

3.86

 worthy

1.81

 worth

1.80

 που

1.58

 which

1.53

of

1.44

so

1.38

 thats

1.34

you

1.32

ที่

1.27

Activations Density 0.164%