INDEX

Explanations

franchise titles and names

The neuron fires strongly on isolated numbers, capitalized words, and title‐style tokens—i.e. it marks section headings, titles, and other header‐like 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

 السياس

-1.02

={`${

-0.94

 선수

-0.87

 jedi

-0.84

ggiata

-0.84

過ぎ

-0.84

まわり

-0.83

 ninjas

-0.83

EventHandler

-0.82

憲法

-0.81

POSITIVE LOGITS

shipping

1.15

bungen

1.09

 junho

1.02

ᜀ

0.99

nically

0.98

AU

0.97

 беже

0.94

verze

0.94

 amino

0.93

Activations Density 0.038%