INDEX

Explanations

introducing specific items or states

This neuron fires strongly on the first word (and sometimes the second) of a new sentence, marking sentence-initial tokens.

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

部分を

-1.44

 denne

-1.36

 واتس

-1.33

amd

-1.28

 ganadores

-1.27

を経

-1.26

 pracy

-1.24

 beide

-1.23

as

-1.23

 chief

-1.23

POSITIVE LOGITS

 menaces

1.52

But

1.29

 conflic

1.28

 kombat

1.27

十几

1.27

 However

1.27

();?>

1.25

 confessions

1.24

 пев

1.23

 calum

1.23

Activations Density 0.071%