INDEX

Explanations

entity or concept followed by specific classifier

This neuron highlights relatively rare or specialized content words—especially named entities, technical terms, and other low‐frequency vocabulary.

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

 which

-2.14

or

-1.96

 when

-1.70

 with

-1.69

for

-1.63

 they

-1.59

not

-1.51

 whether

-1.41

we

-1.33

-1.30

POSITIVE LOGITS

 prób

1.59

uellement

1.57

 człowieka

1.52

begin

1.51

raires

1.50

ricts

1.48

 ausser

1.47

 cortesía

1.46

igable

1.45

 stolet

1.42

Activations Density 0.117%