INDEX

Explanations

intact, complete, preserved

The main thing this neuron does is detect mentions of an item’s physical condition or intactness (e.g., “good condition,” “intact,” “no damage”).

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

-1.43

-1.40

-1.31

-1.30

-1.26

手中

-1.24

-1.22

-1.21

-1.20

POSITIVE LOGITS

alisis

1.38

 intact

1.33

).

1.32

 deteriorating

1.32

otriz

1.25

cyjne

1.24

 courteous

1.23

 although

1.22

قامت

1.21

 categorized

1.20

Activations Density 0.068%