INDEX

Explanations

is not

The neuron strongly activates on common stop‐words—especially linking/copular verbs like “is,” “are,” and auxiliaries such as “have.”

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

potentially

-1.40

both

-1.32

currently

-1.32

BOTH

-1.27

などを

-1.26

definitely

-1.25

neither

-1.24

 facilement

-1.21

 tento

-1.15

 pomys

-1.13

POSITIVE LOGITS

as

1.55

”,

1.24

 reason

1.22

 improves

1.20

 gives

1.20

 suggests

1.16

 requires

1.15

it

1.13

 enhances

1.12

 reduces

1.11

Activations Density 0.075%