INDEX

Explanations

requesting or pleading

The main thing this neuron does is detect instances of pleading or begging language (e.g. “begged,” “pleads,” “plea”).

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

–

-1.44

we

-1.41

’，

-1.39

-1.25

外人

-1.25

ワク

-1.25

三月

-1.23

 analyse

-1.22

 téměř

-1.20

POSITIVE LOGITS

to

2.17

for

1.69

 begging

1.48

 implor

1.43

 telefonu

1.38

 pleading

1.38

 内衣

1.38

 schuldig

1.36

 pleas

1.33

CollectionView

1.33

Activations Density 0.012%