INDEX

Explanations

citation markers

The neuron fires on bracketed numeric or alphanumeric identifiers (e.g. “[pid …]”, error‐code brackets, bibliographic or other ID codes in square brackets).

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

 estadounid

-1.59

 balon

-1.58

 cincin

-1.56

 then

-1.56

 perak

-1.53

 vrijdag

-1.50

 asisten

-1.49

 lâmpada

-1.48

 textil

-1.48

 Añade

-1.48

POSITIVE LOGITS

and

2.16

rasında

1.55

»,

1.51

or

1.50

be

1.43

However

1.41

],

1.35

vs

1.31

 have

1.30

قتل

1.30

Activations Density 0.006%