INDEX

Explanations

still + verb/adjective

The neuron strongly activates on common reporting and ability verbs—especially words like “said,” “able,” “keep,” “present,” and “informed.”

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

鶲

-1.40

teten

-1.38

 guardi

-1.38

 バッジ

-1.38

vartis

-1.34

дного

-1.33

不由得

-1.33

 内衣

-1.31

᎒

-1.29

 vraiment

-1.28

POSITIVE LOGITS

':{

1.46

?"

1.41

’

1.34

 frecuentemente

1.33

!”

1.32

 sebagaimana

1.32

</h3>

1.27

！

1.26

solicited

1.26

!"

1.23

Activations Density 0.038%