INDEX

Explanations

to provide different pictures

The neuron strongly activates on longer, multi-syllable content words (especially less common or more “complex” vocabulary).

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

᩠

-1.84

Răsp

-1.75

)।

-1.62

 atrap

-1.58

 कुछ

-1.57

 händer

-1.54



-1.54

ſtra

-1.52

Č

-1.51

鸰

-1.51

POSITIVE LOGITS

is

1.79

1.77

ዉ

1.69

 than

1.67

 الإلكتر

1.64

1.63

 they

1.59

 других

1.57

of

1.56

 nieuwe

1.52

Activations Density 0.024%