INDEX

Explanations

importantly, most, worst of all

This neuron detects emphatic superlative discourse markers (e.g., “most importantly,” “worst of all,” “most prominently”).

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

 either

-1.84

-1.54

it

-1.52

-1.45

<b>

-1.41

either

-1.38

 because

-1.33

-1.31

-1.30

 those

-1.27

POSITIVE LOGITS

of

1.70

 importantly

1.68

學

1.57

玊

1.52

ciendo

1.52

ным

1.50

новение

1.49

 всех

1.45

 köln

1.45

 制作

1.43

Activations Density 0.029%