INDEX

Explanations

prepositions followed by articles/nouns

The neuron selectively fires on very common short “function” or basic action words (e.g. at, from, add, see, read, get).

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

 вашего

-1.59

 what

-1.58

 quien

-1.56

 даже

-1.53

Уважаемые

-1.49

どのように

-1.48

 eure

-1.45

how

-1.45

見える

-1.45

Только

-1.43

POSITIVE LOGITS

 this

1.83

1.52

hanno

1.50

OF

1.50

Of

1.47

🚒

1.43

 وأ

1.41

 from

1.41

дневно

1.40

と思い

1.39

Activations Density 0.065%