INDEX

Explanations

I would

The neuron fires strongly on first‐person self‐references (e.g. “I,” “my,” “would”), marking subjective statements.

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

 Bescheid

-1.32

：……

-1.31

 again

-1.25

：“……”

-1.22

not

-1.20

が出て

-1.19

 almoço

-1.18

 unfortunately

-1.17

 monaster

-1.16

这也是

-1.16

POSITIVE LOGITS

Даже

1.50

or

1.49

 even

1.41

its

1.36

any

1.35

 même

1.30

Или

1.30

 tôi

1.28

 будь

1.28

ﬂ

1.27

Activations Density 0.064%