INDEX

Explanations

places, currency, questions, or specific terms

The neuron is looking for the structural markers of recipe instructions—namely the list‐item numbers and the punctuation (commas and periods) that separate and end each step.

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

 these

-2.42

 such

-2.23

だけでなく

-1.83

だら

-1.73

 this

-1.67

 zelfs

-1.66

those

-1.65

but

-1.60

だけではなく

-1.59

 ezek

-1.57

POSITIVE LOGITS

了嗎

1.84

地點

1.83

 prefeitura

1.81

Then

1.76

メニューは

1.73

 then

1.70

 your

1.70

いいですね

1.64

зраст

1.63

 before

1.61

Activations Density 0.018%