INDEX

Explanations

strength, power, or quality

The neuron flags words and punctuation that signal quantitative evaluation or degree (e.g. “weak,” “powerful,” “good,” “high,” question marks).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

%")

-1.50

 originalmente

-1.38

%"),

-1.38

 ocup

-1.37

itarias

-1.37

#__

-1.36

"));

-1.32

вля

-1.31

 partid

-1.30

 What

-1.30

POSITIVE LOGITS

 Unglück

1.65

 ewigen

1.43

only

1.30

 Entste

1.30

now

1.30

 geprü

1.27

 soste

1.27

much

1.23

 colorear

1.22

 semplici

1.20

Activations Density 0.017%