INDEX

Explanations

journal titles and academic fields

The neuron fires on journal title abbreviations and other citation‐style metadata in reference lists.

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

為に

-1.77

 that

-1.75

-1.48

of

-1.48

ようです

-1.42

んでしょう

-1.42

ようですが

-1.35

のですね

-1.33

-1.31

を行い

-1.30

POSITIVE LOGITS

がありますが

1.91

Manfaat

1.80

もあって

1.69

や

1.65

は

1.63

Keterangan

1.60

⚌

1.58

Kelebihan

1.54

⛦

1.52

Jangan

1.50

Activations Density 0.011%