INDEX

Explanations

relatively new or painless

The neuron fires on degree-or-hedging modifiers—adverbs and adjectives like “relatively,” “potentially,” “significant,” or “initial” that qualify or soften statements.

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

of

-1.95

 into

-1.56

."

-1.52

 versus

-1.50

 further

-1.39

 bautizo

-1.38

 sorry

-1.34

up

-1.30

 then

-1.26

 them

-1.24

POSITIVE LOGITS

 gått

1.51

☑

1.47

yyyyy

1.47

 fortfarande

1.44

”（

1.41

 Stealth

1.41

 fortsatt

1.35

șit

1.33

šas

1.32

byes

1.32

Activations Density 0.096%