INDEX

Explanations

code snippets and file paths

The neuron detects ASCII‐art–style lines or blocks composed of repeated underscore, pipe, slash, and backslash characters (often used as decorative comment headers or diagrams).

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

-1.45

our

-1.41

-1.39

後

-1.31

𝐎

-1.30

-1.27

’.

-1.26

 caveat

-1.24

 destiné

-1.23

ೢ

-1.22

POSITIVE LOGITS

and

1.40

mampuan

1.33

poň

1.29

设置为

1.29

 produisons

1.24

sew

1.23

Bioaccumulative

1.23

⁧

1.23

Gobierno

1.23

troppo

1.22

Activations Density 0.083%