INDEX

Explanations

code and data separation

This neuron is specialized in detecting literal constants—especially numeric literals and very short string/punctuation tokens—in code.

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

 homicidio

-1.28

ції

-1.25

ený

-1.23

nibus

-1.22

 juſ

-1.18

 Majesté

-1.16

 ส์

-1.16

Luego

-1.15

 его

-1.14

druk

-1.14

POSITIVE LOGITS

or

1.46

 entire

1.37

and

1.37

 whereas

1.23

 procedure

1.21

but

1.20

 this

1.16

 whenever

1.13

 however

1.13

 methodology

1.09

Activations Density 0.135%