INDEX

Explanations

end of section labels

The neuron fires on tokens labeling or linking to a “full summary” section of a document.

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

 their

-1.73

 supremely

-1.51

or

-1.45

:%.

-1.40

֢

-1.38

 every

-1.37

 blatantly

-1.37

 Creating

-1.37

äfts

-1.36

and

-1.34

POSITIVE LOGITS

 alcune

1.96

 alcuni

1.94

があり

1.69

ar

1.65

ara

1.61

で

1.60

le

1.55

などが

1.54

 gigantic

1.53

 diversos

1.52

Activations Density 0.006%