INDEX

Explanations

technical includes and PHP scripts

The neuron activates strongly on words naming concrete entities or formal categories—like organization or product types (e.g. “jet,” “Inc,” “group,” “factory,” “company,” “musical,” “program,” “minister,” “judge”).

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

to

-1.68

 omfatt

-1.64

VERY

-1.60

//{

-1.58

-1.53

 luis

-1.52

THAT

-1.48

睃

-1.44

th

-1.42

 适合

-1.41

POSITIVE LOGITS

all

1.77

new

1.75

 whose

1.60

rassing

1.55

 sonriendo

1.51

or

1.49

 before

1.48

 first

1.45

 every

1.42

 while

1.39

Activations Density 0.239%