INDEX

Explanations

explanation

The neuron strongly activates on prominent content words—proper names, technical terms, and other topic‐specific keywords—in natural, expository (e.g. news‐style) text.

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

ଶ

-1.77

€“

-1.74

⊖

-1.69

佷

-1.68

 '-':

-1.67

 туры

-1.66

髋

-1.63

matory

-1.59

選手の

-1.55

fying

-1.53

POSITIVE LOGITS

can

2.14

 after

1.91

 that

1.70

 häufig

1.67

 souhaitez

1.64

 with

1.61

 before

1.60

 related

1.56

 detergente

1.52

=[];

1.49

Activations Density 0.052%