INDEX

Explanations

concept followed by attribute

The neuron strongly activates on prominent domain-specific nouns—technical or scientific topic keywords (e.g. “body,” “blood,” “telephone,” “text,” “synaptic,” “coding,” “bio,” “enzyme”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

”

-1.86

xically

-1.65

「

-1.64

 reduziert

-1.60

 verhindert

-1.54

 recomiendo

-1.52

 gustaba

-1.51

-1.49

 coût

-1.48

日常

-1.45

POSITIVE LOGITS

and

2.83

 with

1.80

of

1.70

たと

1.57

 sixty

1.52

 eighty

1.52

as

1.50

掲示

1.48

wth

1.44

躊

1.44

Activations Density 0.680%