INDEX

Explanations

religious texts and figures

This neuron activates on mid-word fragments that were split at line breaks (the first chunk of a hyphenated or broken word).

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

־ה

-1.40

ִּ

-1.36

 πρ

-1.29

 dirilis

-1.25

ּוֹ

-1.20

 this

-1.17

иболее

-1.13

our

-1.07

 olisi

-1.04

ابة

-1.03

POSITIVE LOGITS

">[

1.37

>[

1.27

[]){

1.19

LXX

1.16

ὴν

1.12

>();

1.11

**/

1.11

eds

1.10

IAN

1.10

Activations Density 0.005%