INDEX

Explanations

K

The neuron strongly activates on author‐style name tokens (initials and surnames) as seen in academic citation lists.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 romper

0.55

бург

0.55

 burg

0.52

efic

0.52

 मारा

0.51

 अपलोड

0.50

ileno

0.50

godic

0.50

 ስለዚህ

0.50

اره

0.50

POSITIVE LOGITS

KX

0.54

0.53

KF

0.52

矚

0.51

Kwa

0.50

Ｋ

0.49

瞩

0.49

kef

0.49

Cơ

0.48

 умира

0.47

Activations Density 0.226%