INDEX

Explanations

mathematical expressions and sequences

The neuron flags the imperative “give” (and similar prompt words like spelled‐out numbers in instructions) that signal a direct request to provide or compute an answer.

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

and

-1.73

两年

-1.67

archiwizowane

-1.57

-1.52

 seulement

-1.50

-1.44

Wady

-1.44

所有人都

-1.43

 paradise

-1.42

 последних

-1.42

POSITIVE LOGITS

namental

1.85

готовка

1.80

thodon

1.78

nological

1.75

;");

1.71

lised

1.70

chestra

1.69

utenants

1.68

fillment

1.66

aneity

1.66

Activations Density 0.028%