INDEX

Explanations

numbers, percentages, and specific roles

The neuron fires strongly on content‐bearing terms—especially proper names and technical or domain‐specific nouns.

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

高考

-0.98

łe

-0.94

ΡΑ

-0.92

●●

-0.92

ρίες

-0.91

Ошибка

-0.90

 comentários

-0.89

 Sénat

-0.87

دانلود

-0.86

 letto

-0.84

POSITIVE LOGITS

 would

1.23

are

1.17

was

1.12

be

1.08

is

1.02

 have

0.98

 were

0.97

 bởi

0.96

 että

0.95

。

0.95

Activations Density 0.203%