INDEX

Explanations

knowing what the fuck

The neuron flags content‐bearing “keyword” tokens—those distinctive nouns, adjectives, or quantifying terms that signal the main topic or focus of a passage.

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

Same

-1.47

 enfance

-1.25

以下の

-1.20

也在

-1.17

 nouvel

-1.14

新的

-1.13

 также

-1.13

 également

-1.13

SAME

-1.13

рения

-1.11

POSITIVE LOGITS

 each

1.66

 said

1.62

 both

1.45

 there

1.34

its

1.33

 like

1.32

 bowiem

1.31

 that

1.30

↵

1.30

 these

1.29

Activations Density 0.143%