INDEX

Explanations

The neuron strongly activates on “bully/‐ied/‐ing” forms, i.e. words explicitly denoting bullying behavior.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

‌ی

1.59

’

1.35

 তারিখে

1.23

kannya

1.19

霆

1.16

 картины

1.14

sst

1.13

cena

1.12

casa

1.11

 ungroup

1.07

POSITIVE LOGITS

ع

1.98

ส์

1.66

redients

1.47

ly

1.42

ش

1.39

以此

1.33

他

1.31

ا

1.29

不已

1.27

被

1.27

Activations Density 4.173%