INDEX

Explanations

"I" followed by thought or realization

The neuron is triggered by very short, non‐standard “tokens” – namely single characters (like “I”), punctuation marks (commas), and digits – rather than normal multi‐letter 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

in

-1.95

 They

-1.66

get

-1.42

</td>

-1.32

of

-1.31

 Their

-1.26

 With

-1.24

if

-1.21

POSITIVE LOGITS

 použití

1.66

how

1.57

łeś

1.52

 Empfindung

1.49

怎样的

1.47

 způ

1.45

 artesanales

1.45

 mundiales

1.43

 hígado

1.42

 駅前

1.41

Activations Density 0.057%