INDEX

Explanations

can followed by a verb

This neuron specifically fires on occurrences of the modal verb “can.”

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

 then

-1.95

籏

-1.78

也會

-1.73

＞＞

-1.70

 dovr

-1.69

 will

-1.67

？】

-1.66

it

-1.64

 misi

-1.63

 there

-1.61

POSITIVE LOGITS

↵↵

1.63

</h3>

1.59

 unſ

1.54

1.52

 hombro

1.50

that

1.49

is

1.48

 bijvoorbeeld

1.46

疤

1.44

 които

1.43

Activations Density 0.016%