INDEX

Explanations

punctuation followed by Russian

The neuron sharply activates on sentence‐ending punctuation (especially question marks and periods).

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

ook

-1.31

——”

-1.12

 storie

-1.09

にも

-1.08

-1.06

-",

-1.05

 sinds

-1.04

 भी

-1.03

 mivel

-1.02

 deuil

-1.01

POSITIVE LOGITS

and

2.38

but

2.33

And

2.13

And

1.63

why

1.63

 และ

1.56

But

1.55

 maybe

1.52

しかも

1.47

 true

1.30

Activations Density 0.019%