INDEX

Explanations

stay in lieu of

The neuron is primarily detecting the preposition “in” (and closely related locative tokens like “outside,” “overseas,” etc.) marking location‐type phrases.

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

 været

-1.24

 også

-1.18

hrte

-1.18

劻

-1.18

 Also

-1.18

シック

-1.15

harap

-1.14

에도

-1.14

Against

-1.13

 regeringen

-1.13

POSITIVE LOGITS

for

1.95

 instead

1.54

 know

1.37

 when

1.37

how

1.30

“

1.27

 what

1.26

ignal

1.20

nibus

1.18

湾

1.18

Activations Density 0.035%