INDEX

Explanations

methods and means

The neuron fires on words that refer to ways or procedures (e.g. “method,” “methods,” “means,” “preferred”), i.e. mentions of how to do something.

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

 onderdeel

-1.75

 ftate

-1.65

 inmiddels

-1.54

 teď

-1.52

 merkwür

-1.52

 querían

-1.51

 also

-1.45

 belangrijk

-1.42

 betrekking

-1.42

 Sebuah

-1.39

POSITIVE LOGITS

to

1.59

issors

1.56

InetAddress

1.48

</h3>

1.48

ETRIC

1.38

轳

1.38

atschapp

1.38

–

1.37

limus

1.36

 CURLOPT

1.35

Activations Density 0.116%