INDEX

Explanations

castles and palaces

This neuron responds strongly to numeric date tokens (especially years and related numbers) in the text.

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

 delete

-0.77

 Mecca

-0.77

hippo

-0.75

deleting

-0.73

ellite

-0.72

tangent

-0.72

olem

-0.71

 Davos

-0.71

 Wien

-0.70

 $_{\

-0.70

POSITIVE LOGITS

 castle

1.41

 castles

1.30

Castle

1.27

 Castle

1.26

 Castillo

1.03

 castillo

1.03

Château

0.96

 замок

0.92

 قل

0.91

 зам

0.90

Activations Density 0.055%