INDEX

Explanations

of new formation

The neuron selectively activates on abstract process-oriented nouns (nominalizations) such as “development,” “formation,” “generation,” “emission,” and “editing.”

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

or

-1.60

 strives

-1.42

 tries

-1.42

 seeks

-1.39

üstü

-1.37

 what

-1.34

 ensures

-1.34

 puts

-1.34

 trying

-1.32

 longstanding

-1.30

POSITIVE LOGITS



1.58

땃

1.57

only

1.53

 отличаются

1.52

 remarkably

1.50

 somewhat

1.49

 significantly

1.48

 merciless

1.48

 extremely

1.48

 largely

1.48

Activations Density 0.165%