INDEX

Explanations

intensifiers and their effects

This neuron detects emphasis words—intensifying adverbs or adjectives (like “very,” “really,” “quite,” “extremely,” etc.) that amplify sentiment.

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

 schlechte

-1.44

 mittlere

-1.40

 that

-1.39

比べると

-1.35

 можна

-1.26

ituitary

-1.24

 показывает

-1.24

 кто

-1.23

 also

-1.23

 first

-1.20

POSITIVE LOGITS

Bisous

1.48

rophi

1.39

cology

1.38

𖤍

1.37

Bises

1.34

-"

1.34

 ۲۰۰

1.34

 seulement

1.34

bisous

1.34

layed

1.32

Activations Density 0.034%