INDEX

Explanations

PEG, glycol, cobra commands

The neuron fires on technical or domain‐specific identifiers and acronyms—tokens like class names, library names, and scientific abbreviations (e.g. “List,” “Command,” “PEG,” “cobra,” “Exception”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

is

-1.38

 representado

-1.36

🙁

-1.32

护

-1.30

ピュー

-1.25

rugu

-1.23

alians

-1.22

ゴー

-1.22

boutin

-1.22

owie

-1.21

POSITIVE LOGITS

⌦

1.63

 respostas

1.53

is

1.49

 они

1.41

posia

1.40

 グラス

1.38

蘼

1.38

⅘

1.37

 Zitron

1.37

 altes

1.36

Activations Density 0.005%