INDEX

Explanations

files, code, or configurations

The neuron strongly activates on punctuation tokens (especially periods and underscores) and numeric tokens in code- or config-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

 unſ

-1.91

 ſever

-1.88

 които

-1.70

 Anſ

-1.69

 vastly

-1.66

 certaines

-1.64

玙

-1.61

 которые

-1.59

 específica

-1.55

 thuy

-1.53

POSITIVE LOGITS

at

1.80

or

1.63

 will

1.61

one

1.55

Before

1.55

祖国

1.50

as

1.45

1.43

At

1.41

Activations Density 0.063%