INDEX

Explanations

The neuron strongly activates on numeric tokens (digits), punctuation, and uppercase metadata markers—i.e. page numbers, dates, volume/issue labels, wire service tags, and similar document‐header/metadata elements.

numbers and dates

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

 çocuklar

-1.53

 verrou

-1.39

 goû

-1.38

 kanker

-1.38

🔭

-1.36

 réglable

-1.35

𓏸

-1.35

 Chaque

-1.34

 bénévol

-1.31

łgorzata

-1.31

POSITIVE LOGITS

not

1.45

by

1.24

 would

1.23

 five

1.23

six

1.17

at

1.16

of

1.16

no

1.15

be

1.14

 able

1.13

Activations Density 0.005%