INDEX

Explanations

begin statements

The neuron fires on tokens that look like code or markup elements—e.g. programming keywords, LaTeX commands, identifiers, or URL fragments.

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

 usually

-2.27

 also

-1.98

 when

-1.96

但在

-1.87

 should

-1.87

 always

-1.84

 seemed

-1.78

但是在

-1.76

 When

-1.76

↵↵

-1.74

POSITIVE LOGITS

and

1.88

getMenuInflater

1.79

ViewChild

1.74

⠀⠀⠀⠀

1.66

 ſever

1.65

尨

1.65

Ș

1.63

 gegen

1.60

 grudnia

1.56

nischer

1.53

Activations Density 0.017%