INDEX

Explanations

placeholders in questions or code

The neuron activates on isolated non‐alphabetic tokens—numbers, punctuation, and other “junk” symbols.

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

蹀

-1.70

 zoude

-1.68

</h5>

-1.63

 europé

-1.63

genodigd

-1.58

 uniquement

-1.56

谖

-1.55

 poë

-1.52

 perſon

-1.52

Includes

-1.51

POSITIVE LOGITS

in

2.11

 with

1.63

 which

1.60

 from

1.60

and

1.39

 nestled

1.38

 whose

1.36

 heath

1.34

 Vielzahl

1.30

!(:

1.30

Activations Density 0.041%