INDEX

Explanations

temple bars, single engine, square back

The neuron fires on highly structured, technical or specification‐style text—e.g. product/model descriptions, code snippets, FAQs—marked by model numbers, measurements, and other formulaic detail.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 sprá

-0.77

CFS

-0.73

 Larsson

-0.73

 مول

-0.73

diya

-0.72

 Megal

-0.72

 pianos

-0.71

 salivary

-0.70

 képes

-0.69

īga

-0.69

POSITIVE LOGITS

Pri

0.89

 Oakley

0.83

Roo

0.79

 navideño

0.74

Roo

0.72

Tmp

0.71

 Blades

0.70

 blades

0.70

fake

0.69

 Plut

0.69

Activations Density 0.030%