INDEX

Explanations

applied to/in/on

The neuron fires strongly on past-participle verbs (especially in passive constructions) such as “applied,” “formed,” “mounted,” “connected,” “loaded,” “irradiated,” etc.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

жна

-1.52

去年

-1.39

%),

-1.38

????????????????

-1.37

but

-1.37

⁉

-1.35

👀

-1.35

 möjlighet

-1.34

_____________

-1.30

ربی

-1.30

POSITIVE LOGITS

 swarm

1.45

1.43

Somewhat

1.30

1.27

 шесть

1.27

1.26

',

1.23

–

1.20

Activations Density 0.018%