INDEX

Explanations

names followed by common words

The neuron activates on standalone capitalized names—especially people’s surnames and other proper-name 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

apparently

-1.37

個人的には

-1.34

だいたい

-1.33

嵓

-1.30



-1.30

全体的に

-1.28

勿論

-1.26

 besondere

-1.24

そも

-1.23

特に

-1.22

POSITIVE LOGITS

is

1.42

has

1.39

if

1.36

 when

1.27

on

1.26

 will

1.25

you

1.25

 during

1.16

泰

1.14

 with

1.11

Activations Density 0.148%