INDEX

Explanations

code blocks and conditional compilation

The neuron fires on tokens that begin structured code or markup blocks—most notably opening angle-bracket tags (“<”) and key identifiers at the start of code lines (e.g. “themes.add”).

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

なのでしょう

-1.20

 autocollant

-1.19

なのかな

-1.16

なんですね

-1.16

szcz

-1.13

たびに

-1.13

roves

-1.12

を越

-1.09

 manualidades

-1.09

ısını

-1.09

POSITIVE LOGITS

1.50

1.43

1.42

1.39

 only

1.33

1.32

は

1.30

1.23

ちなみに

1.20

1.19

Activations Density 0.007%