INDEX

Explanations

not the

The main thing this neuron does is detect the negation token “not.”

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

and

-1.84

沒有

-1.63

 não

-1.55

ไม่มี

-1.48

 ikke

-1.48

 không

-1.46

Tidak

-1.45

mẫu

-1.43

极其

-1.41

ämpfer

-1.38

POSITIVE LOGITS

 necessarily

2.27

to

1.76

 just

1.70

 because

1.59

 наоборот

1.43

 unlike

1.39

的笑容

1.38

 including

1.33

 only

1.30

The

1.26

Activations Density 0.035%