INDEX

Explanations

access to

The neuron strongly activates on the word “access.”

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

 zostać

-1.50

</td>

-1.34

 trying

-1.29

 oppervlak

-1.20

it

-1.20

 acoged

-1.19

 restauración

-1.18

 pomá

-1.18

 guiding

-1.15

"",

-1.15

POSITIVE LOGITS

of

2.94

to

2.89

𖣔

1.63

 access

1.61

Então

1.60

まさかの

1.59

 Öffnen

1.49

 entonces

1.42

ではでは

1.35

۷

1.34

Activations Density 0.018%