INDEX

Explanations

The neuron fires most strongly on key topic words—especially abstract or evaluative nouns like “qualities,” “skills,” “background,” “strategy,” “implications,” etc.—that label the main subject of a question or discussion.

asking questions about purpose

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

 reje

-1.44

 maquiagem

-1.35

 remun

-1.33

 alug

-1.30

 borboleta

-1.25

 comemor

-1.23

 suspe

-1.23

 bronz

-1.20

 pake

-1.17

 alça

-1.17

POSITIVE LOGITS

or

1.81

to

1.39

in

1.33

 during

1.29

of

1.24

 from

1.18

 españoles

1.16

and

1.13

out

1.12

you

1.09

Activations Density 0.015%