INDEX

Explanations

bullet points and lists

The neuron is selectively detecting LaTeX structural markers—especially left curly‐brace tokens (“{”) that open environments or argument groups.

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

by

-0.94

for

-0.91

 restrictions

-0.85

or

-0.82

 from

-0.81

big

-0.80

Editar

-0.80

 people

-0.79

﹠

-0.79

 talked

-0.78

POSITIVE LOGITS

itemize

1.78

1.50



1.41

 bullet

1.38

bullet

1.36

Bullet

1.30

1.28

bullets

1.27

•

1.25

–

1.23

Activations Density 0.000%