INDEX

Explanations

break the rules and Patterns

This neuron fires strongly on short, standalone lines or headings (often questions or exclamations and quoted snippets), treating them as distinct blocks apart from running text.

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

だが

-1.31

ÈNE

-1.21

 DOLLARS

-1.17

ઃ

-1.07

landschaft

-1.07

 egiten

-1.06

cla

-1.05

ÄT

-1.05

imte

-1.04

IRMED

-1.04

POSITIVE LOGITS

you

1.50

為に

1.34

たぶん

1.34

 çantası

1.25

 évé

1.23

 your

1.22

えっ

1.20

or

1.20

 stratégique

1.18

 dégust

1.15

Activations Density 0.270%