INDEX

Explanations

code keywords and structure

The neuron is triggered by HTML/XML-style tags—especially the “<a name=” anchor tag syntax.

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

 even

-0.75

 schnee

-0.74

 killed

-0.73

ませんでした

-0.72

 retained

-0.71

 adjustable

-0.71

 cleared

-0.70

 Beobach

-0.69

 istor

-0.69

 other

-0.68

POSITIVE LOGITS

ständ

0.86

ATELY

0.85

eavour

0.84

 SOMETHING

0.82

ürlich

0.81

bali

0.80

 клиентов

0.79

woche

0.79

èce

0.77

Activations Density 0.002%