INDEX

Explanations

permission for or denied

The neuron fires on occurrences of “permission” (and its variants like “permissions,” “Permission denied,” etc.), i.e. permission‐related terminology.

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

our

-1.65

 every

-1.64

all

-1.60

 that

-1.60

$}

-1.56

 wollt

-1.56

 time

-1.49

 мы

-1.44

 everybody

-1.44

</th>

-1.39

POSITIVE LOGITS

to

1.52

了下

1.47

 lmao

1.46

Gruss

1.43

KI

1.41

 apenas

1.40

 alcuni

1.40

siehe

1.39

 таких

1.39

following

1.39

Activations Density 0.058%