INDEX

Explanations

articles and specific nouns

The neuron fires on semantically “content‐heavy” words—nonfunction terms like concrete nouns, technical or domain‐specific words, and other contentful vocabulary rather than grammar or stop words.

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

/");

-1.56

ཊ

-1.55

 bouteilles

-1.53

zwi

-1.50

 gethan

-1.49

であった

-1.48

 agonizing

-1.48

 marchandises

-1.47

}/${

-1.45

нились

-1.45

POSITIVE LOGITS

as

1.57

at

1.56

in

1.52

 одна

1.38

 problems

1.38

 sehr

1.27

 began

1.25

в

1.23

 difficulties

1.23

诸多

1.21

Activations Density 0.084%