INDEX

Explanations

multilingual texts about starting/presenting

The neuron fires on the little “header” or metadata labels that introduce definitions or source‐language notes (e.g. tokens like “Latin,” “word,” “office,” etc., at the start of dictionary‐style sections).

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

 manteiga

-1.48

 xícara

-1.38

ENDENT

-1.37

⽐

-1.37

しましたが

-1.32

してます

-1.31

かで

-1.30

噌

-1.28

という感じ

-1.28

なども

-1.27

POSITIVE LOGITS

<strong>

1.48

zustellen

1.45

1.43

or

1.38

 begins

1.38

開始

1.38

of

1.36

 Bentuk

1.33

},

1.31

 naprawdę

1.30

Activations Density 0.003%