INDEX

Explanations

JSON and code structures

This neuron activates on leading whitespace (indentation) in code or data, effectively detecting lines’ indentation levels.

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

of

-2.39

for

-2.08

if

-2.00

 while

-1.60

 this

-1.59

 when

-1.46

-1.39

when

-1.38

 based

-1.27

if

-1.25

POSITIVE LOGITS

 langues

1.19

开玩笑

1.17

1.14

 secuen

1.13

hae

1.09

 volando

1.08

 帅气

1.05

 maintenant

1.05

 nowa

1.05

²,

1.01

Activations Density 0.007%