INDEX

Explanations

names and initials

The neuron fires on capitalized tokens, particularly proper names and acronyms.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 with

-1.57

Зміст

-1.27

 before

-1.27

だが

-1.26

if

-1.26

at

-1.23

 excellent

-1.22

 usato

-1.21

 ujar

-1.21

<0xA3>

-1.20

POSITIVE LOGITS

ちん

1.66

 sembler

1.52

1.48

 didn

1.44

 wouldn

1.41

 wasn

1.41

 tornillos

1.40

 couldn

1.39

乄

1.37

⚣

1.36

Activations Density 0.083%