INDEX

Explanations

proper nouns and programming terms

Looking at the activations, this neuron strongly activates on: - "dwindles" (23.25-38.25) - "include members of" (15.375-31.375) - "agreed" (14.5) - "believe

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

is

-1.80

 créée

-1.39

 vielfach

-1.27

-1.23

 zuneh

-1.20

前は

-1.16

 erinner

-1.16

 using

-1.14

was

-1.13

тнес

-1.13

POSITIVE LOGITS

 theſe

1.53

dáms

1.36

 they

1.33

秾

1.31

 ktorí

1.29

 skriver

1.27

cheidung

1.27

 '/':

1.27

 jedoch

1.26

 Houſe

1.26

Activations Density 0.094%