INDEX

Explanations

used habitually in phrases

The neuron lights up on uncommon content words—especially longer, less frequent vocabulary—rather than on common stopwords.

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

acheteur

-1.67

お話

-1.64

dámské

-1.55

 dừng

-1.52

結局

-1.52

 工房

-1.51

 ceramica

-1.48

historique

-1.48

扨

-1.46

 необхідно

-1.45

POSITIVE LOGITS

ebbe

1.97

↵↵

1.90

 тех

1.69

">//

1.61

1.58

Bissau

1.55

'\\

1.55

/}

1.55

“

1.53

ckeye

1.52

Activations Density 0.074%