INDEX

Explanations

technical terms and punctuation

The neuron fires on words that appear as section or page‐heading labels (e.g. tokens in titles or headings).

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

his

-2.55

its

-2.08

</em>

-2.02

to

-1.89

℣

-1.73

ным

-1.65

-1.63

</h2>

-1.63

 причем

-1.63

 exige

-1.62

POSITIVE LOGITS

’

1.92

’,

1.91

able

1.85

诙

1.77

 hata

1.76

是個

1.75

 hemel

1.72

 hunde

1.70

いてる

1.70

婳

1.69

Activations Density 0.102%