INDEX

Explanations

concepts and findings

The neuron fires on abstract, nominalized nouns—especially those formed with suffixes like “-tion,” “-ance,” or “-ence” (e.g., resistance, determination, defined).

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

 acestea

-1.34

 this

-1.27

なども

-1.20

内容は

-1.19

-1.15

 these

-1.14

それも

-1.14

al

-1.13

-1.10

-1.08

POSITIVE LOGITS

to

1.68

"")

1.42

toit

1.27

 miniaturka

1.27

 tunik

1.27

 Pozna

1.24

 pistolet

1.24

☹

1.23

 konflikt

1.22

katalog

1.20

Activations Density 0.048%