INDEX

Explanations

This neuron does not reliably activate on any code tokens—it remains essentially silent and doesn’t pick out any particular pattern.

comment markers

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

いかがでしたか

-1.56

たまたま

-1.51

明日は

-1.50

マジで

-1.50

びっくり

-1.48

どれだけ

-1.45

さぁ

-1.44

もうすぐ

-1.41

あぁ

-1.40

＊＊＊＊

-1.38

POSITIVE LOGITS

to

1.70

This

1.27

 since

1.23

If

1.20

SRR

1.19

 gese

1.16

 instead

1.16

1.13

 this

1.12

 stadt

1.12

Activations Density 0.007%