INDEX

Explanations

all followed by other words

This neuron activates specifically on the word “all.”

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

それに

-1.43

つまり

-1.36

それを

-1.30

を

-1.30

などが

-1.30

が

-1.27

of

-1.27

があり

-1.25

 Grüßen

-1.22

そして

-1.21

POSITIVE LOGITS

corrhi

1.95

ocarditis

1.63

為に

1.59

ままで

1.48

を行います

1.47

総評

1.46

热血

1.46

€“

1.41

konos

1.40

selves

1.38

Activations Density 0.015%