INDEX

Explanations

question/answer structures

This neuron primarily detects tokens written in all capital letters (e.g., acronyms, emphasized words, or other all-caps 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

and

-1.32

go

-1.30

Usually

-1.29

schuhe

-1.27

 with

-1.24

Almost

-1.20

 pertaining

-1.20

on

-1.20

 शुर

-1.20

 start

-1.19

POSITIVE LOGITS

!!!!!!!

1.44

 !!!!

1.42

 BECAUSE

1.38

biamo

1.37

甚至是

1.31

 !!!!!

1.27

 maravilh

1.27

就行了

1.27

 และ

1.25

 тези

1.23

Activations Density 0.001%