INDEX

Explanations

access to

This neuron detects instances of the word “access,” especially in “access to” constructions indicating gaining or granting entry to something.

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

彼らは

-1.54

you

-1.52

that

-1.47

されてる

-1.42

 ที่

-1.31

詳しい

-1.28

 внутренних

-1.27

If

-1.27

告诉你

-1.27

POSITIVE LOGITS

of

1.95

 icona

1.84

也

1.83

 يول

1.69

 também

1.67

 geval

1.64

'';

1.63

 listop

1.59

 tampoco

1.55

НОВ

1.55

Activations Density 0.040%