INDEX

Explanations

to the / to a

The neuron activates on technical attachment phrases—verbs like “secured,” “hinged,” or “attached” followed by prepositions (“to,” “on”) and articles (“the”) indicating that one component is fastened to another.

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

-1.85

();)

-1.58

:''

-1.55

sm

-1.38

][-

-1.38

JHEP

-1.34

standig

-1.32

-1.30

middot

-1.30

上がり

-1.30

POSITIVE LOGITS

 другой

1.70



1.70

你有

1.67

 Kommentare

1.59

鱘

1.56

my

1.54

ruzione

1.52

薀

1.50

你会

1.48

蟜

1.48

Activations Density 0.062%