INDEX

Explanations

deep investigation or extent

The neuron strongly activates on the token “deep,” effectively detecting occurrences of that specific word.

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

 Also

-2.25

жур

-1.75

</h3>

-1.64

лии

-1.62

-1.61

大腿

-1.60

You

-1.59

も

-1.57

 usual

-1.55

</i>

-1.52

POSITIVE LOGITS

 作り方

1.66

jao

1.63

ativní

1.62

 doubtnut

1.57

 ücretsiz

1.55

퐼

1.55

 любви

1.54

......”

1.52

 suure

1.51

</b>

1.50

Activations Density 0.014%