INDEX

Explanations

answer following punctuation

The neuron fires on the word “answer” (and its morphological variants like “answers,” “answered,” “answering”) whenever it appears.

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

</strong>

-1.59

 even

-1.34

in

-1.34

<>

-1.31

-1.29

-1.25

 blancos

-1.24

’

-1.23

by

-1.20

ساء

-1.18

POSITIVE LOGITS

ες

1.66

྄

1.55

 mediocre

1.45

媗

1.40

䨻

1.38

kać

1.38

谞

1.38

 Voraussetzung

1.34

 fenó

1.33

 molte

1.31

Activations Density 0.064%